main-content

## Über dieses Buch

This book describes the latest research accomplishments, innovations, and visions in the field of robotics as presented at the 13th International Conference on Intelligent Autonomous Systems (IAS), held in Padua in July 2014, by leading researchers, engineers, and practitioners from across the world. The contents amply confirm that robots, machines, and systems are rapidly achieving intelligence and autonomy, mastering more and more capabilities such as mobility and manipulation, sensing and perception, reasoning, and decision making. A wide range of research results and applications are covered, and particular attention is paid to the emerging role of autonomous robots and intelligent systems in industrial production, which reflects their maturity and robustness. The contributions have been selected through a rigorous peer-review process and contain many exciting and visionary ideas that will further galvanize the research community, spurring novel research directions. The series of biennial IAS conferences commenced in 1986 and represents a premiere event in robotics.

## Inhaltsverzeichnis

### Generation of Search Behavior by a Modification of Q-MDP Value Method

We modify Q-MDP value method and observe the behaviors of a robot with the modified method in an environment, where state information of the robot is essentially indefinite. In Q-MDP value method, an action in every time step

is

UEDA, Ryuichi

chosen based on a calculation of expectation values with a probability distribution, which is the output of a probabilistic state estimator. The modified method uses a weighting function with the probability distribution in the calculation so as to give precedence to the states near the goal of the task. We applied our method to a simple robot navigation problem in an incomplete sensor environment. As a result, the method makes the robot take a kind of searching behavior without explicit implementation.

Ryuichi Ueda

### Design of a Low-Cost Vision System for Laser Profilometry Aiding Smart Vehicles Movement

We present a fast and accurate

method

Patruno, Cosimo

to derive the pose of a mobile vehicle moving within bounded paths. A triangulation-based vision system made of a

laser

Marani, Roberto

source, able to generate a line pattern, and a high speed

camera

Nitti, Massimiliano

is applied on the front

side

D’Orazio, Tiziana

of an autonomous vehicle, namely

the

Stella, Ettore

Smoov ASRV platform, which is able to store and retrieve pallets in smart warehouses. The presented system extracts the properties of the emitted laser line on the camera plane and transfers these information to the vehicle reference system. Then, the presence of constitutive landmarks along the path, i.e., holes and bends, permit the estimation of other parameters, such as vehicle speed, enabling the exact control of the vehicle. Further validations have returned accuracies lower than 2 and 3.2 % in distance and tilt measurements with respect to the rail border, respectively.

Cosimo Patruno, Roberto Marani, Massimiliano Nitti, Tiziana D’Orazio, Ettore Stella

### A Comparison of 3D Sensors for Wheeled Mobile Robots

3D sensors are used for many

different

Rauscher, Gerald

applications, e.g., scene reconstruction, object detection, and mobile robots, etc. Several studies on usability and accuracy have been done

for

Dube, Daniel

different sensors. However, all these studies have used different settings for the different sensors. For this

reason

Zell, Andreas

we compare five 3D sensors, including the structured light sensors Microsoft Kinect and ASUS Xtion Pro Live, and the time of flight sensors Fotonic E70P, IFM O3D200 and Nippon Signal FX6, using the same settings. The sensor noise, absolute error, and point detection rates are compared for different depth values, environmental illumination, and different surfaces. Also, simple models of the noise depending on the measured depth are proposed. It is found that structured light sensors are very accurate for close ranges. The time of flight sensors have more noise, but the noise does not increase as strongly with the measured distance. Further, it is found that these sensors can be used for outdoor applications.

Gerald Rauscher, Daniel Dube, Andreas Zell

### Adaptive Model-Based Monitoring for Robots

Continuous and comprehensive monitoring is a key requirement for reliable failure detection. However, the overhead of

the

Kirchner, Dominik

observation

process

Geihs, Kurt

conflicts with the limited resources of a robot platform. Therefore, robot monitoring faces high efficiency requirements. This defines a trade-off between comprehensive observation and monitoring resource overhead. In this paper, we propose an adaptive, model-based monitoring approach that addresses this trade-off. We specify an individual monitoring configuration in an abstract system model to focus the observation on expressive state aspects. Moreover, we introduce adaptivity to further improve the efficiency of the monitoring process. To evaluate this efficiency, we compare our approach with a reference monitoring system. Due to our results, we are confident that the proposed approach significantly reduces the resource overhead.

Dominik Kirchner, Kurt Geihs

### Self-learning RRT* Algorithm for Mobile Robot Motion Planning in Complex Environments

RRT* is a practical and efficient incremental sampling-based

motion

Zhang, Xu

planning algorithm. However, its searching ability is

quite

Lütteke, Felix

inefficient in some cases, due to relying on uniform random sampling like other RRT-based

algorithms

Ziegler, Christian

without taking the environment information and prior knowledge into account, which particularly leads to

many

Franke, Jörg

sampling failures or generation of useless nodes in complex environments. In this paper, we propose an extension of RRT* based on a self-learning strategy and a hybrid-biased sampling scheme to improve the planning efficiency. By taking advantage of the prior knowledge accumulation and cost estimation, the searching tree has higher probability and success rate to extend in difficult areas. We also demonstrate the performance of our algorithm by building some simulation environments for our mobile robot and conclude with the results compared with RRT*.

Xu Zhang, Felix Lütteke, Christian Ziegler, Jörg Franke

### Outdoor Localization of Quad-Rotor Using Extended Kalman Filter and Cell Division Algorithm

This paper proposes a new

technique

Kim, Yoon Ki

that produces the improved local information using low-cost GPS/INS system combined by Extended Kalman Filter and Cell Division

Algorithm

Kim, Ki Jung

the

Lee, Min Cheol

research, the low-cost

GPS

Lee, Jang-Myung

is combined with INS by using extended Kalman filter in order to improve local information. However, this system has the disadvantage that the level of precision for the position information is influenced by the performance of GPS. In order to deal with such disadvantages, the algorithm based on cell division can be adopted. When the quad-rotor flies outdoor, it is possible to predict that its moving path is short, since all the short moving paths of the quad-rotor can be assumed to be straight. Cell division algorithm is used to make such a short moving path and determine the closest local information of the GPS/INS system. Through the above process, an improved kind of local information can be obtained when the quad-rotor flies. Also, the performance of the proposed system can be verified based on various outdoor experiments.

Yoon Ki Kim, Ki Jung Kim, Min Cheol Lee, Jang-Myung Lee

### Cognitive Recognition Under Occlusion for Visually Guided Robotic Errand Service

Naguib, Ahmed M.

A

reliable vision

system

Chen, Xi

for running robotic errand service in an unstructured indoor environment such as homes is difficult to construct.

Many

Lee, Sukhan

visual challenges, such as perspective, clutter, illumination, and occlusion, need to be handled appropriately. While most of previous researches addressed these problems from the contexts of either object recognition or object searching, our proposed approach relies on a solution that combines these two as one. We are proposing a “Cognitive Recognition” System. In the proposed cognitive recognition system, information gathered from scene recognition helps deciding the next optimal perspective, and environmental parameters measurements determine the uncertainty in recognition measurements and thus the proper probability map update used in object search. We show particularly in this paper how this approach provides a practical solution to cluttered and occluded environments. And we demonstrate the results with our HomeMate Service Robot.

Ahmed M. Naguib, Xi Chen, Sukhan Lee

### An Application of Laser-Based Autonomous Navigation for Data-Center Monitoring

Data-center monitoring has been a critical subject of research in recent years. Mobile robots have been successfully employed in the industrial field to efficiently perform common tasks. In this paper, we

report

Rosa, Stefano

some preliminary results on the study and development of a robotic system, in which a mobile robot

equipped

Russo, Ludovico Orlando

with a laser range sensor

and

Bona, Basilio

an Inertial Measurement Unit (IMU) is able to autonomously navigate in a data-center

room

Marco, Gaspardone

for accurate monitoring of critical measurements, such as servers’

external

Farulla, Giuseppe Airó

temperature, humidity and other physical quantities. The

robot

Carlone, Luca

is able to autonomously create a map of a previously unknown room, localize therein and execute a list of

measurements

Antonini, Roberto

at different locations, which are provided by the user via a web graphical user interface (GUI). The robot is able to find the best trajectory to reach the given locations, while avoiding static and moving obstacles. The particular characteristics of the data-center scenario introduce specific problems related to map creation and localization using laser-based techniques (e.g., irregular surfaces as metal grids and high symmetry of the environment), which must be properly taken into account and are discussed throughout the paper. Preliminary experimental results show that the system is able to create a consistent map of the environment, to correctly localize itself therein and to follow a given path.

Stefano Rosa, Ludovico Orlando Russo, Giuseppe Airó Farulla, Luca Carlone, Roberto Antonini, Gaspardone Marco, Basilio Bona

### Semantic Topological Map-Based Smart Wheelchair Navigation System for Low Throughput Interface

This paper presents a new navigation

system

Wei, Zhixuan

designed for the smart wheelchair, which interacts with a human

user

Chen, Weidong

using a low throughput interface. Through the low throughput interface, the user can only

give

Wang, Jingchuan

a very limited number of command

types

Wang, Huiyu

with a low frequency, meaning that a low throughput interface requests the user to do more steps, taking a longer time, just to select the desired goal. In order to decrease the number of operations requested to the user, our navigation system is designed to refine the alternative targets by selecting the significant targets and organize them in a binary tree for reducing user’s operation. We also introduce a user-friendly visual feedback to display to the user in order to show the current state and the prompt commands to the user, too. This navigation system is successfully tested in a real environment.

Zhixuan Wei, Weidong Chen, Jingchuan Wang, Huiyu Wang

### Robust Supervisory-Based Control Strategy for Mobile Robot Navigation

This work introduces a novel control strategy to allow a class of

mobile

Furci, Michele

robots to robustly navigate in a dynamic and potentially cluttered environment. The proposed approach combines a high-level motion planner, designed

considering

Naldi, Roberto

the supervisory control theory, and a low-level stabilizing feedback control law. Taking advantage of a

symbolic

Paoli, Andrea

description of the vehicle dynamics

and

MArconi, Lorenzo

of the environment, the supervisor reactively selects the current motion primitive to be executed so as to reach the desired target location optimally with respect to a given index cost. Sufficient conditions ensuring boundedness of the tracking error are derived in order to handle the interaction between the discrete-time dynamics of the supervisor and the continuous-time dynamics of the low-level control loop in charge of tracking the desired reference. The resulting approach allows to employ supervisory control tools online without affecting the stability properties of the continuous-time low-level control loop. The results are demonstrated by considering, as application, the kinematic model of an aerial vehicle navigating in a cluttered environment.

Michele Furci, Roberto Naldi, Andrea Paoli, Lorenzo Marconi

### Vision-Based Hybrid Map Building for Mobile Robot Navigation

A hybrid mapping framework

is

Üzer, Ferit

presented in this work. The goal is to obtain better computational efficiency than pure metrical mapping techniques and better accuracy as well as

usability

Korrapati, Hemanth

for robot guidance and navigation compared to

the

Royer, Eric

topological mapping. Image

sequences

Mezouar, Youcef

acquired in an

environment

Lee, Sukhan

by manually driving a robot are used to build a hierarchical map representation by using an image sequence partitioning (ISP) technique that uses local image features. The hierarchical map built can be understood as a topological map with nodes corresponding to certain regions in the environment. Each node in turn is made up of a set of images acquired in that region. These maps are further augmented with metrical information at those nodes which correspond to image subsequences acquired while the robot is turning as a part of its trajectory. Metrical information becomes invaluable during autonomous robot navigation through these places. Hence, we call the resulting maps hybrid since they primarily contain topological information and metrical information at places that are important for navigation. Experimental results obtained on a sequence acquired in an outdoor environment are provided to demonstrate our approach.

Ferit Üzer, Hemanth Korrapati, Eric Royer, Youcef Mezouar, Sukhan Lee

### pRoPhEt MAS: Reactive Planning Engine for Multiagent Systems

Autonomous mobile robots can

substantially

Saur, Daniel

increase their effectiveness in dynamic environments using planning. This paper proposes a design for a soft real-time planning system for autonomous robots and offers a generic and modular approach to control a team of robots. Our system pRoPhEt MAS is based on ALICA (A Language for Interactive Cooperative Agents) and offers the coordination of team behaviors at runtime. In the evaluation scenario the system pRoPhEt MAS uses a

state

Geihs, Kurt

of the art planner “Fast Downward Planning System.” The evaluation focuses on planning during execution time. The team executes the best solutions found, selected by the heuristic, under certain time constraints. The results show that the execution with soft real-time planning is as good as sequential planning and execution. Hence, it offers the ability to react quickly in dynamic domains.

Daniel Saur, Kurt Geihs

### Autonomous Determination of Locations for Observing Home Environment Using a Mobile Robot

The paper focuses on determining locations in a typical home environment for monitoring or observing the surroundings using an indoor mobile robot. Currently, these locations are manually selected for the robots. In the process of

autonomously

Mehdi, Syed Atif

evaluating minimum number of key locations, the proposed methodology targets free spaces in the environment

that

Berns, Karsten

may provide maximum observability to a mobile robot with limited range of sensor systems. The technique also ensures that these locations are at a distance from obstacles in the environment in order to guarantee sufficient space for robot navigation. The experiments have been performed both in real apartment using an autonomous mobile robot and in simulation with a variety of environments. The results demonstrate an area coverage of up to

$$96\,\%$$

96

%

with minimal locations computed in fairly acceptable time. These locations can be used in various scenarios like monitoring an elderly person in the home environment.

Syed Atif Mehdi, Karsten Berns

### Enhanced Stochastic Mobility Prediction on Unstructured Terrain Using Multi-output Gaussian Processes

Outdoor robots such as

planetary

Lui, Sin Ting

rovers must be able to navigate safely and reliably in order to successfully

perform

Peynot, Thierry

missions in remote or hostile environments. Mobility prediction is critical to achieving this goal

due

Fitch, Robert

to the inherent control uncertainty faced by robots traversing natural terrain. We propose a novel algorithm for

stochastic

Sukkarieh, Salah

mobility prediction based on multi-output Gaussian process regression. Our algorithm considers the correlation between heading and distance uncertainty and provides a predictive model that can easily be exploited by motion planning algorithms. We evaluate our method experimentally and report results from over 30 trials in a Mars-analogue environment that demonstrate the effectiveness of our method and illustrate the importance of mobility prediction in navigating challenging terrain.

Sin Ting Lui, Thierry Peynot, Robert Fitch, Salah Sukkarieh

### Multi-modal Sensors Path Merging

In this paper, we propose an original method to merge maps from different robots. Each map was built using a camera which can be different: perspective, fish-eye, or omnidirectional. Each robot creates its own local map, while the main goal is to build a global map assuming that the paths overlap each other on at least one segment of the path. The first step is to find this common part by using a trajectory correlation method. Then the rigid transformation between trajectories is computed and used to merge paths. Since the robot paths are not sufficient to determine if the matching is correct, an image sensor is required do finish the procedure.

Léo Baudouin, Youcef Mezouar, Omar Ait-Aider, Helder Araújo

### Continuous Motion Planning for Service Robots with Multiresolution in Time

We present an approach to continuous motion planning with multiresolution in time. Our approach is based on stochastic trajectory optimization for motion planning (STOMP) and designed to decrease the optimization time in order to enable frequent replanning. Since service robots operate in

environments

Steffens, Ricarda

with dynamic obstacles, it is likely that planned trajectories become invalid over time. Thus, it is not

necessary

Nieuwenhuisen, Matthias

to provide trajectories with a uniform high resolution. Our multiresolutional

approach

Behnke, Sven

implicitly considers the uncertainty of the future by providing a trajectory with a gradually coarser schedule, which is refined trough replanning. In addition to employing temporal multiresolution, we speed up trajectory optimization by initializing replanning with the previous plan. The proposed multiresolution STOMP is evaluated in simulation in comparison to the original STOMP implementation. Our experiments show that multiresolution STOMP reduces the planning time and, hence, is able to avoid dynamic obstacles.

Ricarda Steffens, Matthias Nieuwenhuisen, Sven Behnke

This paper describes the construction of roadmaps for mobile robots

using

Matamoros, Mauricio

vector quantization clustering of the free space. A RGB-D camera is

used

Savage, Jesus

to find

planes

Rivera, Carlos

that

represent

Bermudez, Alejandro

objects in

the

Contreras, Luis

environment as

well

Negrete, Marco

as walls and floors. The free space is found by separating the objects’ and walls’

planes

Figueroa, Israel

from the floor’s plane, that represents the space where the mobile robot navigates. Vector quantization

technique

Pacheco, Abel

is used to partition the free space into regions, and the centroids of the regions become the nodes of a roadmap, that is used by a mobile robot to navigate.

Jesus Savage, Luis Contreras, Israel Figueroa, Abel Pacheco, Alejandro Bermudez, Marco Negrete, Mauricio Matamoros, Carlos Rivera

### Scan Planning for RGB-D Mapping Based on Grid Map

Autonomously building 3D

maps

Cheng, Min

is important for robots in real-world applications. Even the state-of-the-art RGB-D mapping

techniques

Wang, Feng

rely on

manually

Chen, Xiaoping

collected image data. It is confronted with a great challenge since: (1) the distance and viewing angle of the data provided by RGB-D cameras are too limited, (2) and indoor scenes are too large compared to the valid data range of RGB-D cameras. In this paper, we develop a novel method to generate the scan plans toward the problem using 2D grid maps as priori knowledge and taking into account the drawbacks of RGB-D cameras and the limitations of RGB-D mapping systems. Scan plans allow robots to scan indoor environments autonomously with relevant grid maps. An implementation of this method is applied on our real robot, and proven to be robust and effective in experiments.

Min Cheng, Feng Wang, Xiaoping Chen

### On-Road Trajectory Planning for General Autonomous Driving with Enhanced Tunability

In order to achieve smooth

autonomous

Gu, Tianyu

driving in real-life urban and highway environments, a motion planner must generate trajectories that are locally smooth and responsive (reactive), and at the same time, far-sighted and intelligent (deliberative). Prior approaches achieved

both

John M., Dolan

planning qualities for full-speed-range operations at a high computational cost. Moreover, the planning formulations were mostly a

trajectory

Lee, Jin-Woo

search problem based on a single weighted cost, which became hard to tune and highly scenario-constrained due to overfitting. In this paper, a pipelined (phased) framework with tunable planning modules is proposed for general on-road motion planning to reduce the computational overhead and improve the tunability of the planner.

Tianyu Gu, John M. Dolan, Jin-Woo Lee

### Curve-Graph Odometry: Removing the Orientation in Loop Closure Optimisation Problems

In robot odometry and SLAM applications the real trajectory is estimated incrementally. This produces an accumulation of errors which gives raise to a drift in the trajectory. When revisiting a previous position this drift becomes observable and thus it can be corrected by applying loop closing techniques. Ultimately a loop closing

process

Gutiérrez-Gómez, Daniel

leads to an optimisation problem where new constraints between poses obtained from loop detection are applied to the

initial

Guerrero, J.J.

incremental estimate of the trajectory. Typically this optimisation is jointly applied on the position and orientation of each pose of the robot using the state-of-the-art pose-graph optimisation scheme on the manifold of the rigid body motions. In this paper, we propose to address the loop closure problem using only the positions and thus removing the orientations from the optimisation vector. The novelty in our approach is that, instead of treating trajectory as a set of poses, we look at it as a curve in its pure mathematical meaning. We define an observation function which computes the estimate of one constraint in a local reference frame using only the robot positions. Our proposed method is compared against state-of-the-art pose-graph optimisation algorithms in 2 and 3 dimensions. The main advantages of our method are the elimination of the need of mixing the orientation and position in the optimisation and the savings in computational cost due to the reduction of the dimension of the optimisation vector.

Daniel Gutiérrez-Gómez, J. J. Guerrero

### Toward Localizing both Static and Non-static RFID Tags with a Mobile Robot

Recent research shows more and

more

Liu, Ran

interest in exploring the mapping task of tagged-items in the context of inventory with

mobile

Zell, Andreas

robots in passive radio-frequency identification (RFID)-equipped infrastructures. However, mapping RFID tags is quite challenging, since the characteristics of radio signals are heavily influenced by environmental effects (e.g., reflection, diffraction, or absorption). This paper presents the augmented particle filter, which is able to recover from mapping failures of static RFID tags and localize non-static RFID tags. Furthermore, although negative information is usually considered to be less informative than positive information, we exploit the usefulness of negative information for RFID-based mapping. We show that a careful examination of negative information improves the mapping accuracy and helps to recover from mapping failures and relocalize non-static RFID tags. Additionally, we compare the particle filter-based approach to our previous grid-based Markov localization approach. Last, we demonstrate a mobile system, which is able to approach both static and non-static RFID tags, and avoid obstacles at the same time.

Ran Liu, Andreas Zell

### RRS: Rapidly-Exploring Random Snakes a New Method for Mobile Robot Path Planning

Recently, sampling-based path planning

algorithms

Baizid, K.

have been

implemented

Chellali, R.

in many practical

robotics

Luza, R.

improvements

Vitezslav, B.

have been dedicated to

the

Arrichiello, F.

returned solution (quality) and sampling process. The aim of this paper is to introduce a new technique that improves the classical rapidly-exploring random trees (RRT) algorithm. First, the sampling step is modified in order to increase the number of possible solutions in the free space. Second, within the possible solutions, we apply an optimization scheme that gives the best solution in term of safety and shortness. The proposed solution, namely, rapidly-exploring random snakes (RRS) is a combination of a modified deformable Active Contours Model (called Snakes) and the RRT. The RRS takes the advantage of both RRT and deformable Snakes contours, respectively, in: rapidly searching new candidate nodes in the free space and circumnavigating obstacles by calculating a safe sub-path in the free space toward the new node created by the RRT. In comparison to the classical RRT, the proposed algorithm increases the

probability of completeness

, accelerates the convergence and generates a much safer and shorter open-loop solution, hence, increasing considerably the efficiency of the classical RRT. The proposed approach has been validated via numerical simulations and experimental results with a mobile robot.

K. Baizid, R. Chellali, R. Luza, B. Vitezslav, F. Arrichiello

### Layered Mission and Path Planning for MAV Navigation with Partial Environment Knowledge

Successful operation of micro aerial vehicles in partially known environments requires

globally

Nieuwenhuisen, Matthias

consistent plans based on incomplete environment models and quick reactions to unknown obstacles by

means

Behnke, Sven

of real-time planning of collision-free trajectories. In this paper, we propose a complete layered mission and navigation planning system based on coarse prior knowledge and local maps from omnidirectional onboard obstacle perception. We generate trajectories in a multilayered approach: from mission planning to global and local trajectory planning to motion control.

Matthias Nieuwenhuisen, Sven Behnke

### Multi-Vehicle Adaptive Planning with Online Estimated Cost Due to Disturbance Forces

This paper proposes an adaptive planning architecture for multivehicle teams subject to an uncertain, spatially varying disturbance force. Motivated by a

persistent

Desaraju, Vishnu R.

surveillance task, the planning architecture is designed with three hierarchical levels. The highest level

generates

Liu, Lantao

interference-free routes for the entire team to monitor areas of

interest

Michael, Nathan

that have higher uncertainty. The lower level planners compute trajectories that can be tracked accurately along these routes by anticipating the effects of the disturbance force. To this end, the vehicles maintain an online estimate of the disturbance force, which drives adaptation at all planning levels. A set of simulation results validate the proposed method and demonstrate its utility for persistent surveillance.

Vishnu R. Desaraju, Lantao Liu, Nathan Michael

### A Generalized Extended Kalman Filter Implementation for the Robot Operating System

Accurate state estimation for a mobile

robot

Moore, Thomas

often requires the fusion of data from multiple sensors. Software that performs sensor fusion should therefore support the

inclusion

Stouch, Daniel

of a wide array of heterogeneous sensors. This paper presents a software package,

robot_localization

, for the robot operating system (ROS). The package currently contains an implementation of an extended Kalman filter (EKF). It can support an unlimited number of inputs from multiple sensor types, and allows users to customize which sensor data fields are fused with the current state estimate. In this work, we motivate our design decisions, discuss implementation details, and provide results from real-world tests.

Thomas Moore, Daniel Stouch

### Stereo Graph-SLAM for Autonomous Underwater Vehicles

The increasing use of

Autonomous Underwater Vehicles

(AUV) in industrial or scientific applications makes the vehicle

localization

Negre Carrasco, Pep Lluis

one of the challenging questions to consider for the mission success. Graph-SLAM has emerged as a promising

approach

Bonin-Font, Francisco

in land vehicles; however, due to the complexity of the aquatic media, these systems have been rarely applied in underwater vehicles. The few existing approaches are focused

on

Codina, Gabriel Oliver

very particular applications and require important amounts of computational resources, since they optimize the coordinates of the external landmarks and the vehicle trajectory, all together. This paper presents a simplified and fast general approach for stereo graph-SLAM, which optimizes the vehicle trajectory, treating the features out of the graph. Experiments with robots in aquatic environments show how the localization approach is effective underwater, online at 10 fps, and with very limited errors. The implementation has been uploaded to a public repository, being available for the whole scientific community.

Pep Lluis Negre Carrasco, Francisco Bonin-Font, Gabriel Oliver Codina

### Robust Onboard Visual SLAM for Autonomous MAVs

This paper presents a

visual

Yang, Shaowu

simultaneous localization and mapping (SLAM) system consisting of a robust visual odometry and an efficient back-end

with

Scherer, Sebastian A.

loop-closure detection and pose-graph optimization (PGO). Robustness of the visual odometry is achieved by utilizing dual cameras pointing

different

Zell, Andreas

directions with no overlap in their respective fields of view mounted on an micro aerial vehicle (MAV). The theory behind this dual-camera visual odometry can be easily extended to applications with multiple cameras. The back-end of the SLAM system maintains a keyframe-based global map, which is used for loop-closure detection. An adaptive-window PGO method is proposed to refine keyframe poses of the global map and thus correct pose drift that is inherent in the visual odometry. The position of each map point is then refined implicitly due to its relative representation to its source keyframe. We demonstrate the efficiency of the proposed visual SLAM algorithm for applications onboard MAVs in experiments with both autonomous and manual flights. The pose tracking results are compared with the ground truth data provided by an external tracking system.

Shaowu Yang, Sebastian A. Scherer, Andreas Zell

### Exploiting Structural Properties of Buildings Towards General Semantic Mapping Systems

Semantic mapping is one of the most

active

Luperto, Matteo

and promising research areas within autonomous mobile robotics. Informally, a

semantic

Amigoni, Francesco

map associates a high-level human-understandable label (like “office” or “corridor”) to a portion of an environment. Most semantic mapping approaches are based on classifiers that, given some features perceived by robot sensors in a physical place, associate a semantic label to the place. These approaches are often tested on a limited number of homogeneous places (e.g., few rooms within a single building). This line of action seems to hinder the development of methods for constructing semantic maps that can be (re)used in a number of previously unseen environments. In this paper, we aim at contributing to make semantic mapping methods more general. In particular, we focus on indoor environments and we consider the following research question: to what extent are the semantic mapping approaches shown to label rooms in a single building expected to work when applied to different buildings?

Matteo Luperto, Francesco Amigoni

### A Kinect-Based Front-End for Graph-SLAM Using Plane Matching in Planar Indoor Environments

We present a pose graph optimization approach which enables a mobile robot to create 3D maps of planar indoor environments using a Microsoft Kinect camera. Rather than using

point

Yuan, Zehui

features, the approach relies on the registration of planar surfaces. Plane matching is used to implement a (

front-end

) for the

construction

Rosa, Stefano

of a pose graph, which is then optimized by a state-of-the-art (

back-end

). Vertical planes are extracted from

acquired

Russo, Ludovico

point clouds associated

to

Bona, Basilio

the poses of the robot; then a plane matching algorithm is used to create constraints among successive robot poses. Place revisiting episodes are detected using 3D features in order to provide loop-closing constraints. These constraints provide the input for a pose graph optimization algorithm, which computes an estimate of the robot trajectory. Finally, the 3D map is created by attaching to each pose of the trajectory the corresponding planes. Planar surfaces are more robust and descriptive with respect to point features and provide an accurate estimate of rotations. Moreover, the front-end combines geometric and appearance-based information to filter out outliers and perform robust plane association. Preliminary experimental results in real environments show that the approach is able to create 3D maps which are consistent and close to reality.

Zehui Yuan, Stefano Rosa, Ludovico Russo, Basilio Bona

### Landmark Rating and Selection for SLAM in Dynamic Environments

Goal-oriented acting in

dynamic

Hochdorfer, Siegfried

environments is a challenging task for a mobile robot. A fundamental

problem

Neumann, Heiko

to be solved is to map the environment during exploration. Since everyday, environments are typically

not

Schlegel, Christian

static, landmarks can occur and disappear at any time. Therefore, a SLAM approach must be able to cope with the characteristics of such environments. This work presents a multicriteria utility function to select landmarks for SLAM in dynamic environments. The landmark utility function takes into account the salience, the probability of reobservation, and the relevance for localization of a landmark. Taking into account these criteria, now enables the selection of landmarks for SLAM in dynamic environments. The performance of the approach is shown in a real-world experiment with a P3DX-platform in a living room environment.

Siegfried Hochdorfer, Heiko Neumann, Christian Schlegel

### Fast Active SLAM for Accurate and Complete Coverage Mapping of Unknown Environments

In this paper, we

present

Lenac, Kruno

an active SLAM solution with an active loop closing component which is independent on exploration component and at the same time allows high accuracy robot’s pose estimation and complete environment mapping. Inputs to our SLAM algorithm are RGBD image from the Kinect sensor and odometry estimates obtained from inertial measurement unit and wheel encoders. SLAM is based on the exactly sparse delayed state filter for real-time estimation of robot’s trajectory, vision-based

pose

Kitanov, Andrej

registration, and loop closing. The active component ensures that localization remains accurate over a long period of time by sending the robot to close

loops

Maurović, Ivan

if a criterion function satisfies the predefined value. Our criterion function depends on the number of states predicted without an update between predictions,

information

Dakulović, Marija

gained from loop closing and the sheer distance between the loop closing state location and the current

robot

Petrović, Ivan

location. Once a state in which a loop closure should occur is reached and an update is performed, the robot returns to its previous goals. Since the active component is independent on the exploration part, the SLAM solution described in this paper can easily be merged with any existing exploration algorithm and the only requirement is that the exploration algorithm is able to stop exploration at any time and continue the exploration after the loop closing was accomplished. In this paper, we propose an active SLAM integration with the 2D laser range finder-based exploration algorithm that ensures the complete coverage of a polygonal environment and therefore a detailed mapping. The developed Active SLAM solution was verified through experiments which demonstrated its capability to work in real-time and to consistently map polygonal environments.

Kruno Lenac, Andrej Kitanov, Ivan Maurović, Marija Dakulović, Ivan Petrović

### Local Multi-resolution Surfel Grids for MAV Motion Estimation and 3D Mapping

For autonomous navigation in restricted environments, micro aerial vehicles (MAV) need to create 3D maps of their surroundings and must track their motion within these maps. In this paper, we

propose

Droeschel, David

an approach to simultaneous localization and mapping that is based

on

Stückler, Jörg

the measurements of a lightweight 3D laser-range finder. We aggregate laser-range

measurements

Behnke, Sven

by registering sparse 3D scans with a local multiresolution surfel map that has high resolution in the vicinity of the MAV and coarser resolutions with increasing distance, which corresponds well to measurement density and accuracy of our sensor. Modeling measurement distributions within voxels by surface elements allows for efficient and accurate registration of 3D scans with the local map. The incrementally built local dense 3D maps of nearby key poses are registered globally by graph optimization. This yields a globally consistent dense 3D map of the environment. Continuous registration of local maps with the global map allows for tracking the 6D MAV pose in real time. In experiments, we demonstrate accuracy and efficiency of our approach.

David Droeschel, Jörg Stückler, Sven Behnke

### Voronoi-Based Heuristic for Nonholonomic Search-Based Path Planning

This paper proposes the

use

Wang, Qi

of a Voronoi-based heuristic to significantly speed up search-based nonholonomic path planning. Using generalized Voronoi diagrams (GVD) and in this manner exploiting geometric information about the obstacles, the

presented

Wulfmeier, Markus

approach is able to considerably reduce computation time while satisfying differential constraints using

motion

Wagner, Bernardo

primitives for exploration. A key advantage compared to the common use of Euclidean heuristics is the inherent ability to avoid local minima of the cost function, which can be caused by, e.g., concave obstacles. Therefore, the application of the Voronoi-based heuristic is particularly beneficial in densely cluttered environments.

Qi Wang, Markus Wulfmeier, Bernardo Wagner

### 3D Pose Detection for Articulated Vehicles

The knowledge about relative poses within a tractor/trailer

combination

Fuchs, Christian

is a vital prerequisite for kinematic modeling and trajectory estimation. In case of autonomous vehicles or

driver-assistance

Zöbel, Dieter

systems, for example, the monitoring of an attached passive trailer is

crucial

Paulus, Dietrich

for operational safety. A

3-D

pose detection sensor based on an optical approach suitable for uneven ground is presented and evaluated against a

2-D

Christian Fuchs, Dieter Zöbel, Dietrich Paulus

### Construction of 3D Environment Models by Fusing Ground and Aerial Lidar Point Cloud Data

A lot of research work deals with the building of 3D environment models, e.g. by lidar-based 6D SLAM on

ground

Langerwisch, Marco

vehicles. Because these single vehicle approaches always are afflicted by partial occlusion of the environment, we

propose

Krämer, Marc Steven

to fuse point cloud data taken by ground and aerial vehicles. Therefore, we use

manually

Kuhnert, Klaus-Dieter

steered ground and

aerial

Wagner, Bernardo

vehicles equipped with localization sensors and laser scanners to record point cloud data. The point cloud data is fused predominantly by existing state-of-the-art algorithms and data formats in ROS. Finally, Octomaps are calculated as common environment models. Two real world experiments in structured and unstructured outdoor environments are presented. The resulting point clouds and maps are evaluated qualitatively and quantitatively.

Marco Langerwisch, Marc Steven Krämer, Klaus-Dieter Kuhnert, Bernardo Wagner

### Fast Classification of Accessible Terrain with a 3D Laser Range Finder

Identification of drivable paths is the ability for navigation in rural environments with minimal infrastructure.

Timo Röhling

### Data Handling in Large-Scale Surface Reconstruction

Using high resolution laser scanners, it is possible to create consistent 3D point

clouds

Wiemann, Thomas

of large outdoor environments in a short time. Mobile systems are able to measure whole cities efficiently and

collect

Mrozinski, Marcel

billions of data points. Such large amounts of data can usually not be

processed

Feldschnieders, Dominik

on a mobile system. One approach to create a feasible environment representation that can be used on mobile

robots

Lingemann, Kai

is to compute a compact polygonal environment representation. This paper addresses problems and solutions

when

Hertzberg, Joachim

processing large point clouds for surface reconstruction.

Thomas Wiemann, Marcel Mrozinski, Dominik Feldschnieders, Kai Lingemann, Joachim Hertzberg

### Vision-Based Person Detection for Safe Navigation of Commercial Vehicle

Vision-Based solution of safe

Piao, Songlin

for commercial vehicles with fish-eye camera is presented in this paper. This work aims to develop a system which is able to detect persons or objects around

commercial

Berns, Karsten

vehicles for preventing any accidents. This is achieved by integrating classifier-based window searching algorithm and ego-motion-based algorithm into the system. The classifier is trained using cascaded support vector machine and ego-motion is estimated only based on captured images from camera. Test results show that designed system provides high detection rate when it is applied in commercial vehicle.

Songlin Piao, Karsten Berns

### Leveraging Open Data for Supporting a Cloud Robotics Service in a Smart City Environment

At the dawn of cloud

robotics

Ermacora, G.

, one of the biggest challenges is to successfully exploit the power offered by the

internet

Toma, A.

for acquiring and sharing information, in order to build a common knowledge base among the agents. In this

paper

Antonini, Roberto

, we propose a cloud robotics service for emergency management in a smart city scenario. The

application

Rosa, Stefano

is capable of collecting significant open data over the internet. Then it uses them in the mission planning phase, in order to autonomously define a set of waypoints toward the target coordinates.

G. Ermacora, A. Toma, R. Antonini, S. Rosa

### Throughput Optimisation in Ad Hoc Networks of Communication-Aware Mobile Robots

We study throughput optimisation of ad hoc

networks

Ngo, Trung Dung

of communication-aware mobile robots. The mobile robots equipped with sensing and communication capacities can maintain connectivities and estimate the quality of communication links with their neighbouring peers. The mobile robots self-organise a wireless ad hoc network for transmitting environment exploited data from sources to destinations. Graph-based network model and artificial potential force-based connectivity maintenance are integrated in different ways for the control design of mobile robots. We consider throughput optimisation in twofold: (1) routing-aware optimisation and (2) routing-unaware optimisation. The Monte Carlo simulation results are comparatively analysed and discussed according to the performance metrics.

Trung Dung Ngo

### On the Formation Control of Multiagent Systems Under Nearly Cyclic Pursuit

We consider the problem of formation

control

of multiple agents under the nearly cyclic pursuit strategy. A control law is designed under

this

Ngo, Trung Dung

strategy for

n

agents to rendezvous at a point and also form a specific formation, dictated by an extra agent, called beacon. We derive constraint on the gain in

k

circulant

matrix to make the system stable. Formation tracking for

n

agents is achieved by applying an exogenous input to the beacon. We prove that the agents in nearly cyclic interaction are controllable under the beacon. Simulations and analytical results demonstrate the effectiveness of the proposed method.

### Leveraging Area Bounds Information for Autonomous Multirobot Exploration

In this paper, we

propose

Liu, Tsung-Ming

an approach, the Space-Based Potential Field (SBPF) approach, to control multiple robots for area exploration missions that focus on robot dispersion. The SBPF method is based on a potential field approach that

leverages

Lyons, Damian M.

knowledge of the overall bounds of the area to be explored. This additional information allows a simpler potential field control strategy for all robots but which nonetheless has good dispersion and overlap performance in all the multirobot scenarios while avoiding potential minima. Both simulation and robot experimental results are presented as evidence.

Tsung-Ming Liu, Damian M. Lyons

### Bio-inspired Classification in the Architecture of Situated Agents

Cognitive development

concerns

Gini, G.

the evolution of human mental capabilities through experience earned during life. Important

features

Franchi, A.M.

needed to

accomplish

Mutti, F.

this target are the self-generation of motivations and goals as

well

Ferrini, F.

as the development of complex behaviors consistent with these goals. Our

target

Manzotti, R.

is to build such a bio-inspired

cognitive

Gallo, F.

architecture for

situated

Mutti, F.

agents, capable of integrating new sensing data from any source. Based on neuroscience assessed concepts, as neural plasticity and neural coding, we show how a categorization module built on cascading classifiers is able to interpret different sensing data. Moreover, we see how to give a biological interpretation to our classification model using the winner-take-all paradigm.

G. Gini, A. M. Franchi, F. Ferrini, F. Gallo, F. Mutti, R. Manzotti

### Adaptive Multiagent Path Planning with Dynamic Heuristic

Multiagent path planning is a challenging

problem

Behbooei, Mohammed

in robotics. Basically, there are two types of approaches on this problem. Centralized approaches in which by some state space searching algorithms obtaining an optimal

solution

MohaimenianPour, SeyedMehdi

is guaranteed however this completeness concludes on drawbacks like exponential

time

Ghidary, Saeed Shiry

and space complexity. On the contrary, incompleteness and nonoptimality in decentralized approaches result in polynomial time complexity; therefore, we face a trade-off between completeness and time complexity. We propose a complete centralized semi-coupled algorithm including three different phases which uses discrete time stamps to search in state space to find the optimal solution. The first phase is called Advanced-Solver based on A* search, the second phase is called Easy-Solver which tries to complete Advanced-Solvers solution faster and optimally, and finally in the third phase in order to reduce time complexity, we introduce a deadlock-handler heuristic which prevents opening some useless states by pruning state space tree. Our algorithm has been implemented on our simulator and the result was also tested on real robots.

SeyedMehdi MohaimenianPour, Mohammed Behbooei, Saeed Shiry Ghidary

### Communication-Aware Route Selection in Wireless Sensor Networks

We consider the problem of optimal

route

selection for wireless sensor network in the presence of path loss, multipath fading, interference, and environmental noise. The communication-aware

route

Hung, Pham Duy

selection strategy is proposed by incorporating realistic communication model portraying the underlying dynamics of wireless link. The link quality is

characterized

Tuah, Nor Jaidi

by the probability of successfully received packets over a communication link, so-called reception probability. We utilize reception probability as a

metric

Ngo, Trung Dung

for communication quality-oriented route selection and to compare its performance with the conventional metrics, i.e., Hop count and Euclidean distance. The simulation results demonstrate that reception probability-based route selection provides optimal end-to-end throughput in wireless sensor networks.

Adnan Fida, Pham Duy Hung, Nor Jaidi Tuah, Trung Dung Ngo

### A Scalable, Decentralised Large-Scale Network of Mobile Robots for Multi-target Tracking

A scalable, decentralised large-scale

network

Hung, Pham Duy

of mobile robots for multi-target tracking is addressed in this paper. The decentralised

control

Vinh, Tran Quang

is originally built up by behavioural control but upgraded with decentralised robot

control

Ngo, Trung Dung

for connectivity maintenance and decentralised connectivity control for hierarchical connectivity removal, allowing the network expansion for tracking and occupying spatially distributed targets. The multi-target tracking algorithm guarantees that the mobile robots reach targets at very high efficiency, while at least an interconnectivity network connecting all the mobile robots is preserved for information exchange. The Monte Carlo simulation results illustrate characteristics of the decentralised control as well as its scalability through several experimental scenarios.

Pham Duy Hung, Tran Quang Vinh, Trung Dung Ngo

### Global Reconfiguration of a Team of Networked Mobile Robots Among Obstacles

This paper presents a full system

demonstration

Fitch, Robert

of dynamic sensor-based reconfiguration of a networked robot team. Robots sense

obstacles

Alempijevic, Alen

in their environment locally and dynamically adapt their global geometric configuration to

conform

Peynot, Thierry

to an abstract goal shape. We present a novel two-layer planning and control algorithm for team reconfiguration that is decentralised and assumes local (neighbour-to-neighbour) communication only. The approach is designed to be resource-efficient and we show experiments using a team of nine mobile robots with modest computation, communication and sensing. The robots use acoustic beacons for localisation and can sense obstacles in their local neighbourhood using IR sensors. Our results demonstrate globally specified reconfiguration from local information in a real robot network, and highlight limitations of standard mesh networks in implementing decentralised algorithms.

Robert Fitch, Alen Alempijevic, Thierry Peynot

### Automatic Detection and Feature Estimation of Windows from Mobile Terrestrial LiDAR Data

This work presents a new

method

of automatic window detection in 3D LiDAR point clouds obtained from mobile terrestrial data acquisition systems in the

urban

Checchin, Paul

environment. The proposed method first segments out 3D points belonging to the building façade from the 3D urban point cloud and

then

Trassoudaine, Laurent

projects them onto a 2D plane parallel to the building façade. After point inversion within a watertight boundary, windows are segmented out based on geometrical information. The window features are then estimated exploiting both symmetrically corresponding windows in the façade as well as temporally corresponding windows in successive passages, based on ANOVA measurements. This unique fusion of information not only accommodates for lack of symmetry but also helps complete missing features due to occlusions. The estimated windows are then used to refine the 3D point cloud of the building façade. The results, evaluated on real data using different standard evaluation metrics, demonstrate the efficacy of the method.

Ahmad K. Aijazi, Paul Checchin, Laurent Trassoudaine

### Practically Identifiable Model of Robotic Manipulator for Calibration in Real Industrial Environment

problem

Klimchik, Alexandr

of robotic manipulator calibration in

real

Caro, Stephane

industrial environment.

Particular

Furet, Benoit

attention is

paid

Pashkevich, Anatol

to the practical identifiability of the model parameters, which completely differs from the theoretical one that relies on the rank of the observation matrix only, without taking into account essential differences in the model parameter magnitudes and the measurement noise impact. To solve the problem, several model reduction methods are proposed. The advantages of the developed approach are illustrated by an application example that deals with the geometric calibration of an industrial robot used in aerospace industry.

Alexandr Klimchik, Stephane Caro, Benoit Furet, Anatol Pashkevich

### 3D Mapping by a Robotic Fish with Two Mechanical Scanning Sonars

3D mapping is one of the most

significant

Chen, Ling

abilities for autonomous underwater vehicles (AUV). This paper proposes a 3D mapping

algorithm

Wang, Sen

for a robotic fish using two mechanical scanning sonars (MSSs) with one being forward-looking

and

Hu, Huosheng

the other downward-looking. Combined with inertial measurement unit (IMU), the forward-looking MSS is used for 2D SLAM (simultaneous localization and mapping)

by

Ryuh, Young-sun

which the 2D poses of the vehicle are optimally obtained by applying a pose-based GraphSLAM. Based on the

estimated

Yang, Gi-Hun

2D poses, depth and orientation, the measurements from the downward-looking sonar are used to build the 3D map by adapting 3D mapping algorithm Octomap while taking into account the pose uncertainty. The effectiveness of the proposed algorithm is verified by extensive simulations which also show that it can generate more informative 3D map than the scenario where no uncertainty of poses is considered.

Ling Chen, Sen Wang, Huosheng Hu, Young-sun Ryuh, Gi-Hun Yang

### Specific Person Detection and Tracking by a Mobile Robot Using 3D LIDAR and ESPAR Antenna

Tracking a specific

person

Misu, Kazuki

is one of the important tasks of mobile service robots. This paper describes a novel and reliable strategy for detecting and

tracking

Miura, Jun

a specific person by a mobile robot in outdoor environments. The robot uses 3D LIDARs for person detection and identification, and a directivity-variable antenna (called

ESPAR antenna

) for locating a specific person even under occluded and/or out-of-view situations. A sensor fusion approach realizes a reliable tracking of a specific person in actual outdoor environments. Experimental results show the effectiveness of the proposed strategy.

Kazuki Misu, Jun Miura

### Automatic Extraction of Structural Representations of Environments

Robots need a suitable

representation

Capobianco, Roberto

of the

surrounding

Gemignani, Guglielmo

world to operate in a structured but dynamic environment. State-of-the-art

approaches

Bloisi, Domenico Daniele

usually rely on a combination of metric and topological maps and require an expert to

provide

Nardi, Daniele

the knowledge to the robot in a suitable format. Therefore, additional symbolic knowledge cannot

be

Iocchi, Luca

easily added to the representation in an incremental manner. This work deals with the problem of effectively binding together the high-level semantic information with the low-level knowledge represented in the metric map by introducing an intermediate grid-based representation. In order to demonstrate its effectiveness, the proposed approach has been experimentally validated on different kinds of environments.

Roberto Capobianco, Guglielmo Gemignani, Domenico Daniele Bloisi, Daniele Nardi, Luca Iocchi

### Generic Distance-Invariant Features for Detecting People with Walking Aid in 2D Laser Range Data

People detection in 2D laser

range

Weinrich, Christoph

data is a popular cue for person tracking in mobile robotics. Many approaches are designed to detect

pairs

Wengefeld, Tim

of legs. These approaches perform well in many public environments. However, we are working on an assistance robot for stroke patients in a rehabilitation center, where most of the people need walking aids. These tools occlude or touch the legs of the patients. Thereby,

approaches

Volkhardt, Michael

based on pure leg detection fail. The essential contribution of this paper are generic distance-invariant

range

Scheidig, Andrea

scan features for

people

Gross, Horst-Michael

detection in 2D laser range data. The proposed approach was used to train classifiers for detecting people without walking aids, people with walkers, people in wheelchairs, and people with crutches. By the use of these features, the detection accuracy of people without walking aids increased from an

$$F_1$$

F

1

score of 0.85 to 0.96, compared to the state-of-the-art features of Arras et al. Moreover, people with walkers are detected with an

$$F_1$$

F

1

score of 0.95 and people in wheelchairs with an

$$F_1$$

F

1

score of 0.94. The proposed detection algorithm takes on average less then 1 % of the resources of a 2.8 GHz CPU core to process 270

$$^{\circ }$$

laser range data with an update rate of 12 Hz.

Christoph Weinrich, Tim Wengefeld, Michael Volkhardt, Andrea Scheidig, Horst-Michael Gross

### Statistical Modelling of Object Detection in Stereo Vision-Based Driver Assistance

In this work, a statistical

analysis

Stellet, Jan Erik

of object

detection

Schumacher, Jan

for stereo vision-based driver assistance systems is presented. Analytic modelling has not been attempted

previously

Lange, Oliver

due to the complexity of dense disparity maps and state-of-the-art algorithms. To approach this problem, a simplified algorithm for object detection in stereo

images

Branz, Wolfgang

which allows

studying

Niewels, Frank

error propagation is considered. In order to model the input densities, vehicle contours are approximated by Gaussian Mixture Models and distance dependent measurement noise is taken

into

Zöllner, J. Marius

account. Theoretical results are verified with Monte Carlo methods and real-world image sequences. Using the proposed model, a prediction on the uncertainty in object location and optimal threshold selection can be obtained.

Jan Erik Stellet, Jan Schumacher, Oliver Lange, Wolfgang Branz, Frank Niewels, J. Marius Zöllner

### Optical Kinematics of Dual Mirror Laser Rangefinders

The topic of kinematics of laser rangefinders has received little attention in the robotics literature, even though such sensors have been the

perception

Kelly, Alonzo

sensors of choice on commercial AGVs, field robots, and aerial robots for some time. Unlike the physical mechanisms employed in manipulators and mobile robots, the basic operation for bending the optical path of the transmit and receive beams of a laser rangefinder is often a reflection about a mirror, and this operation cannot be modeled naturally as a rotation. In recognition of the uniqueness of optical reflection mechanisms, this paper presents a formulation for modeling 2 axis scanning laser rangefinders based on a matrix reflection operator.

Alonzo Kelly

### Real-Time Face Recognition Using Local Ternary Patterns with Collaborative Representation-Based Classification for Mobile Robots

The ability to recognize faces is a crucial element for human-robot interaction. In this paper, we present an algorithm for mobile robots to detect, track, and recognize human faces accurately, even when humans go through different illumination conditions. We track faces using a tracker that combines the algorithm of an adaptive correlation filter with a Viola-Jones object detection. This tracker adapts to scale changes and to rotation of the face, and its occlusion. It also adapts to complex changes of background and illumination. Recognizing the tracked face is established by using an algorithm that combines local ternary patterns and collaborative representation based classification (CRC). This combination enhances the efficiency of face recognition under different illumination and noisy conditions. Our method achieves high recognition rates on challenging face databases and can run in real time on mobile robots.

Duc My Vo, Andreas Zell

### Theta-Disparity: An Efficient Representation of the 3D Scene Structure

We propose a new

representation

Nalpantidis, Lazaros

of 3D scene structure, named theta-disparity. The proposed representation is a 2D

angular

Kragic, Danica

depth

histogram

Kostavelis, Ioannis

that is calculated

using

Gasteratos, Antonios

a disparity map. It models the structure of the prominent objects in the scene and reveals their radial distribution relative to a point of interest. The proposed representation is analyzed and used as a basic attention mechanism to autonomously resolve two different robotic scenarios. The method is efficient due to the low computational complexity. We show that the method can be successfully used for the planning of different tasks in the industrial and service robotics domains, e.g., object grasping, manipulation, plane extraction, path detection, and obstacle avoidance.

Lazaros Nalpantidis, Danica Kragic, Ioannis Kostavelis, Antonios Gasteratos

### Unsupervised Range Image Segmentation and Object Recognition Using Feature Proximity and Markov Random Field

In this paper, we propose a

framework

Rizzini, Dario Lodi

for unsupervised range image segmentation and object recognition that exploits feature similarity and

proximity

Oleari, Fabio

as leading criteria in the processing steps. Feature vectors are distinctive traits like color, texture and shape of the regions of the scene; proximity of similar features enforces

classification

Atti, Andrea

and association decisions. Segmentation is performed by dividing the input

point

Aleotti, Jacopo

cloud into voxels, by extracting and clustering features from each voxel, and by refining such

segmentation

Caselli, Stefano

through Markov Random Field model. Candidate objects are selected from the resulting regions of interest and compared with the models contained in a dataset. Object recognition is performed by aligning the models with the refined point cloud clusters. Experiments show the consistency of the segmentation algorithm as well as the potential for recognition even when partial views of the object are available.

Dario Lodi Rizzini, Fabio Oleari, Andrea Atti, Jacopo Aleotti, Stefano Caselli

### Twist-State Classifier for Floating Marine Biomass Based on Physical Simulation

This paper describes new approaches for classifying twist of seaweeds. There are no evaluation measures of the twist formation of complicated objects quantitatively because the definition of a qualitative twist is a difficult problem. The twist of seaweeds is one of these problems. In this paper, we propose three factors (physical, geometric, and time factor of twist) for characterizing the twist state, and we develop the twist-state classifier based on these factors. Additionally, the analysis experiment verifies how the classifier shows the classification accuracy of twist state.

Jun Ogawa, Hiroyuki Iizuka, Masahito Yamamoto, Masashi Furukawa

### Localization of Unmanned Aerial Vehicles Using Terrain Classification from Aerial Images

In this paper we investigate the benefit of terrain classification for self-localization of a flying robot. The key idea is to use aerial images, which are already available from online databases such as GoogleMaps™, as reference map and to match images taken with a downward looking camera with this map. Using different terrain classes as features, we can make sure that our method is invariant to lighting/weather changes as well as seasonal variations or minor changes in the environment. A particle filter is used to register the query image with parts of the map. The proposed method has shown to work on image data from both simulated and real flights.

Andreas Masselli, Richard Hanten, Andreas Zell

### Kinect@Home: A Crowdsourced RGB-D Dataset

Algorithms for 3D localization, mapping, and reconstruction are getting increasingly mature. It is time to also make the datasets on which they are tested more realistic to reflect the conditions in the homes of real people. Today algorithms are tested on data gathered in the lab or at best in a few places, and almost always by the people that designed the algorithm. In this paper, we present the first RGB-D dataset from the crowdsourced data collection project Kinect@Home and perform an initial analysis of it. The dataset contains 54 recordings with a total of approximately 45 min of RGB-D video. We present a comparison of two different pose estimation methods, the Kinfu algorithm and a key point-based method, to show how this dataset can be used even though it is lacking ground truth. In addition, the analysis highlights the different characteristics and error modes of the two methods and shows how challenging data from the real world is.

Rasmus Göransson, Alper Aydemir, Patric Jensfelt

### Object Recognition and Tracking for Indoor Robots Using an RGB-D Sensor

In this paper, we extend and generalize our previously published approach on RGB-D based fruit recognition to be able to recognize different kinds of objects in front of our mobile system. We therefore first extend our segmentation to use depth filtering and clustering with a watershed algorithm on the depth data to detect the target to be recognized. We forward the processed data to extract RGB-D descriptors that are used to recoup complementary object information for the classification and recognition task. After having detected the object once, we apply a simple tracking method to reduce the object search space and the computational load through frequent detection queries. The proposed method is evaluated using the

random forest

(RF) classifier. Experimental results highlight the effectiveness as well as real-time suitability of the proposed extensions for our mobile system based on real RGB-D data.

Lixing Jiang, Artur Koch, Andreas Zell

### Recognition of Three-Dimensional Branch Structure and Fruits Identification in a Tree Based on It

This paper describes a method to recognize a branch structure of a fruit tree. We can identify fruits and

branches

Yumoto, Yuya

using the branch structure. Identification of them enables us to gather data of

each

Mizuuchi, Ikuo

fruit and branch. Collected data can be utilized for growing management and sales. We describe a method to obtain three-dimensional branch structure from the point cloud. We succeeded in recognizing a simple dummy fruit tree. We propose a method to stably recognize the same branch structure. We also tested the recognizing algorithm using a real fruit tree (a persimmon tree). We could recognize correct branches from point clouds that did not have an occlusion area and unnecessary points such as leaves.

Yuya Yumoto, Ikuo Mizuuchi

### 3D Object Recognition Using Convolutional Neural Networks with Transfer Learning Between Input Channels

RGB-D data is getting ever more

interest

Alexandre, Luís A.

from the research community as both cheap cameras appear in the market and the applications of this type of data become more common. A current trend in processing image data is the use of convolutional neural networks (CNNs) that have consistently beat competition in most benchmark data sets. In this paper, we investigate the possibility of transferring knowledge between CNNs when processing RGB-D data with the goal of both improving accuracy and reducing training time. We present experiments that show that our proposed approach can achieve both these goals.

Luís A. Alexandre

### Moving Objects Tracking on the Unit Sphere Using a Multiple-Camera System on a Mobile Robot

Detection and tracking of moving objects with camera systems mounted on a mobile robot presents a formidable problem since the ego-motion of the robot

and

Ćesić, Josip

the moving objects jointly form a challengingly discernible motion in the image. In this paper, we are concerned with multiple-camera systems, namely the Ladybug

$$^{\textregistered }2$$

®

2

camera, whose

perspective

Marković, Ivan

images were used to detect motion and subsequently perform the tracking of multiple objects on the sphere. This enabled us to account for the continuity of the scene

which

Petrović, Ivan

is achieved by the sensor in an image stitching process on the sphere. The objects are tracked on the sphere with a Bayesian filter based on the von Mises–Fisher distribution and the data association is achieved by the global nearest neighbor method, for which the distance matrix is constructed by deriving the Rényi

$$\alpha$$

α

-divergence for the von Mises–Fisher distribution. The prospects of the method are tested on a synthetic and real-world data experiments.

Josip Ćesić, Ivan Marković, Ivan Petrović

### Visual Mapping in Light-Crowded Indoor Environments

Due to the recent success of affordable RGBD cameras,

solutions

Klüssendorff, Jan Helge

to the Visual Simultaneous Localization and Mapping (VSLAM) problem has experienced a huge leap. To

enable

Ehlers, Kristian

accurate

mapping

Maehle, Erik

solutions, most of the proposed solutions expect static environments. Thinking of industrial applications, there is no guarantee for static environments. The SLAM algorithm has to cope with moving objects like human beings. We present an approach to detect moving objects in RGBD camera images. The approach is based on point cloud and image filtering techniques. We present test results using publicly available datasets. We further show the performance and influence of the algorithm on mapping and on the accuracy of a visual SLAM system.

Jan Helge Klüssendorff, Kristian Ehlers, Erik Maehle

### Nonparametric Calibration for Depth Sensors

In this paper, we propose a quick and easy approach to estimate the undistortion

function

Cicco, Maurilio Di

of RGBD sensors. Our method does not rely on the knowledge of the sensor model, on the use of a specific calibration pattern or on external SLAM systems to track

the

Iocchi, Luca

device position. We compute a nonparametric approximation of the undistortion function by applying regression methods to calibration data that can be

acquired

Grisetti, Giorgio

wherever a sufficiently large planar surface is observed. The procedure is fast, easy, and be used on-line. Experimental results show a significant improvement when using undistorted images in applications like mapping.

Maurilio Di Cicco, Luca Iocchi, Giorgio Grisetti

### An Architecture for Telenoid Robot as Empathic Conversational Android Companion for Elderly People

In Human-Humanoid Interaction (HHI), empathy is the

crucial

Sorbello, Rosario

key in order to overcome the current limitations of social robots. In facts, a

principal

Chella, Antonio

defining characteristic of human social behaviour is empathy. The present

paper

Giardina, Marcello

presents a robotic architecture for an android robot as a basis for natural empathic human-android interaction. We start from the hypothesis that the robots, in order to become

personal

Nishio, Shuichi

companions need to know how to empathic interact with human beings. To validate our research, we have used the proposed system with the minimalistic humanoid robot Telenoid. We have conducted human-robot interactions

test

Ishiguro, Hiroshi

with elderly people with no prior interaction experience with robot. During the experiment, elderly persons engaged a stimulated conversation with the humanoid robot. Our goal is to overcome the state of loneliness of elderly people using this minimalistic humanoid robot capable to exhibit a dialogue similar to what usually happens in real life between human beings. The experimental results have shown a humanoid robotic system capable to exhibit a natural and empathic interaction and conversation with a human user.

Rosario Sorbello, Antonio Chella, Marcello Giardina, Shuichi Nishio, Hiroshi Ishiguro

### Designing Robots for Positive Communication with Senior Citizens

Robots for elderly care are meant to

increase

Kuwamura,Kaiko

the quality of care services. We designed a teleoperated android called Telenoid to support communication with the elderly. At first sight, Telenoid’s odd appearance often provokes negative reactions from people of various ages. After they hug and interact with it, their impressions become more positive. However, experiments in Japan and

Denmark

Nishio, Shuichi

suggest that the

elderly

Ishiguro, Hiroshi

easily accept Telenoid from their first interactions with it, but the reason remains unexplained. This paper hypothesizes and discusses possible reasons why Telenoid is accepted by the elderly. We think that Telenoid triggers and enhances the ability of the elderly to imagine and positively complete the information, thus making them feel attracted to it. Based on this hypothesis, we discuss the factors that trigger the imagination and lets the senior citizens form a positive impression of the robot.

Kaiko Kuwamura, Shuichi Nishio, Hiroshi Ishiguro

### Online Interaction of a Human Supervisor with Multi-Robot Task Allocation

In this paper, an approach is presented that allows a human supervisor to efficiently interact with task allocation in a multi-robot team (MRTA). The interaction is based on online

modification

Kurowski, Karen

of the setting of the employed MRTA optimization algorithm during its computation. For the example of a computationally expensive mixed-integer linear programming

algorithm

Stryk, Oskar von

it is demonstrated how to achieve up to optimal solution quality, while simultaneously reducing the required calculation time compared to a fully autonomous optimization. The supervisor is enabled to rate feasible, intermediate solutions based on objective or subjective quality criteria and personal expertise. In that way, also suboptimal solutions can be chosen to be satisfactory, and the solver can be terminated without the need to wait for the completion of the computation of the optimal solution. An event-based communication concept with queries is used as an efficient means of implementation of the interaction. Furthermore, the supervisor can support the MRTA solver in finding good solutions by defining crucial parts of the solution structure. These intuitive commands are internally translated into constraints and are added to the problem as lazy constraints. This combination of human expertise and state-of-the-art optimization algorithms allows to achieve up to potentially optimal task allocation in much shorter time.

Karen Kurowski, Oskar von Stryk

### Human-Robot Collaborative Remote Object Search

Object search is one of the typical tasks for remotely controlled service robots. Although object recognition technologies have been well developed, an

efficient

Miura, Jun

search strategy (or viewpoint planning method) is still an issue. This paper describes a new approach to human-robot

collaborative

remote object search. An analogy for our approach is

ride on shoulders

; a user controls a fish-eye camera on a remote robot to change views and search for a target object,

independently

Chikaarashi, Kota

of the robot. Combined with a certain level of automatic search capability of the robot, this collaboration can

realize

Sugiyama, Junichi

an efficient target object search. We developed an experimental system to show the feasibility of the approach.

Jun Miura, Shin Kadekawa, Kota Chikaarashi, Junichi Sugiyama

### Gathering and Conceptualizing Plan-Based Robot Activity Experiences

Learning from experiences is an

effective

approach to enhance robot’s competence. This paper focuses on developing capabilities for a robot to obtain robot activity experiences and

conceptualize

Lim, Gi Hyun

the experiences as plan schemata, which are used as heuristics for the robot to make plans in similar situations. The plan-based robot activity

experiences

Lopes, Luís Seabra

are obtained through human-robot interactions where a teaching action from a command-line user

interface

Pinho, Armando J.

triggers recording of an experience. To represent human-robot interaction activities, ontologies for experiences and user instructions are integrated into a robot ontology. Recorded experiences are episodic descriptions of the robot’s activities including relevant perceptions of the environment, the goals pursued, successes, and failures. Since the amount of experience data is large, a graph simplification algorithm based on

ego networks

is investigated to filter out irrelevant information in an experience. Finally, an approach to robot activity conceptualization based on deductive generalization and abstraction is presented. The proposed system was demonstrated in a scenario where a PR2 robot is taught how to “serve a coffee” to a guest, in the EU project RACE.

Vahid Mokhtari, Gi Hyun Lim, Luís Seabra Lopes, Armando J. Pinho

### Considerate Behavior of Robots Based on Individual’s Preference

Adaptive behaviors by a robot could result in a mistake in human life, since the robot does not know what the user wants exactly. We think that the robot should learn individual’s preference of the user before executing adaptive behaviors. As a result, it is expected that the robot can execute adaptive behaviors appropriate to the user. We defined this as “considerate behaviors.” In order to let the robot execute considerate behaviors, we use a neural network to teach individual’s preference and considerate behaviors to the robot. Furthermore, we find the effective way of selecting supervisors of neural network to make the robot execute considerate behaviors with a few selected supervisors. We tested it in simulation and also in real life with a real robot named DARwIn-OP.

Jeonghun Baek, Ikuo Mizuuchi

### Gait Measurement for Human Behavior Estimation Against Autonomous Mobile Robot

To realize a safe collision

avoidance

Yorozu, Ayanori

of autonomous mobile robots, understanding human behavior against the robots is important. This paper has proposed a gait

measurement

Takahashi, Masaki

method for human behavior estimation against autonomous mobile robot. The proposed method using a laser range sensor consists of five observed leg patterns recognition and global nearest neighbor (GNN)-based data association with a variable validation region based on the state of each leg. To verify the effectiveness of the proposed system, a verification test in a hospital that staff is recruited as participants were carried out. From the experimental results in the hospital, we confirm that the proposed method can reduce the chance of losing track of both legs and the variable validation region can reduce the chance of false tracking.

Ayanori Yorozu, Masaki Takahashi

### Optimizing Support Vector Machine with Genetic Algorithm for Capacitive Sensing-Based Locomotion Mode Recognition

Capacitive sensing has been

proven

Song, Yi

valid for locomotion mode recognition as an alternative of popular electromyography-based methods in the control of powered prostheses. In

order

Zhu, Yating

to obtain higher recognition accuracy, in this paper, we try to improve the support vector machine (SVM)-based classifier by

selecting

Zheng, Enhao

suitable kernel

function

Tao, Fei

and optimizing the

parameters

Wang, Qining

with genetic algorithm (GA). According to different phases of the gait, the phase-dependant GA-SVM models are built and the recognition accuracy increase from 94.0 to

$$99.1\,\%$$

99.1

%

, which is satisfactory for practical applications.

Yi Song, Yating Zhu, Enhao Zheng, Fei Tao, Qining Wang

### A Tag-Based Recommender System

Recommender systems are being used more and more on the web thanks to their ability to predict user preferences and drive user attention

toward

Caro, Pietro De

new items, increasing sales, and engagement. However, the use of such systems is still very limited to e-commerce and music or movies websites and, most of the times, the user is presented with recommendations limited to products. Our idea is to provide suggestions that are content-agnostic and that can be used to recommend mixed types

of

Pini, Maria Silvia

contents at the same time (for example, images, posts, and products). In such a way, the

power

Sambo, Francesco

of recommender systems can be exploited in very diverse contexts using a unique model with few adjustments. To achieve this, we provide a tag-based recommender system with a highly scalable implementation that is proposed with the aim of providing performance and reusability in a Software as a Service (SaaS) package.

Pietro De Caro, Maria Silvia Pini, Francesco Sambo

### Clustering of Humanoid Robot Motions Executed in Response to Touch

We perform a study on

responses

Basoeki, Fransiska

that should be performed by robots when touched by humans. To study the kinds of

robot

DallaLibera, Fabio

responses in general, there is a need for clustering similar responses. We present the use of Multiple correspondence analysis (MCA) and hierarchical clustering

method

Menegatti, Emanuele

as a way of

clustering

Pagello, Enrico

different humanoid robot postures. MCA is commonly used to analyze data with discrete variables. Since

clustering

Ishiguro, Hiroshi

solely by MCA was impractical for our application, hierarchical clustering method was performed to aid the direct inspection of the robot response clusters.

Fransiska Basoeki, Fabio Dalla Libera, Emanuele Menegatti, Enrico Pagello, Hiroshi Ishiguro

### Artificial Hand with Stiffness Adjuster

The paper deals with a five-finger hand, which is based on the original

finger

Koganezawa, K.

mechanism consisting of a planetary gear system and the compound four bar linkages. It takes an all-in-one design: all of the actuators (total five DC motors) are

embedded

Ito, A.

into a palm, while finger parts have no electronic devices for attaining to be inherently safe as an end-effector. The mechanism allows us adaptive synergic motions of three joints of a finger (MP, PIP, and DIP) according to the shape of the objects to be gripped. The hand has a novel mechanism for adjusting stiffness of fingers, which provides an ability to give passive gripping force to a gripping object according to its elasticity. Driving tests show that it achieves fundamental motions of a human hand in daily life without any sensory feedback and also shows that the stiffness adjuster works effectively.

K. Koganezawa, A. Ito

### Hierarchical Task Networks as Domain-Specific Language for Planning Surgical Interventions

The following paper addresses the challenges of defining surgical workflows. Surgical workflows have to deal with medical and technical aspects on different

levels

Schreiter, Luzie

of abstraction in order to ensure safety. We propose hierarchical task networks (HTN) as a unifying domain-specific language (DSL) for the definition of surgical workflows. The DSL describes relations and

dependencies

Bihlmaier, Andreas

in state

sequences

Raczkowsky, Jörg

and surgical actions for complex workflows on varying levels of detail. With an HTN planner we are able to decompose high-level steps into primitive actions and identify

all

Wörn, Heinz

possible workflows together with their paths through the intervention. This information can be used to identify missing or inaccurate information in literature and consequently improve the workflow and safety of the surgical intervention. By means of a case study we present a detailed HTN-based DSL for

Laparoscopic Cholecystectomy

to show the advantage of using our particular approach to workflow modeling.

Andreas Bihlmaier, Luzie Schreiter, Jörg Raczkowsky, Heinz Wörn

### EndoSnake—A Single-Arm-Multi-Port MIRS System with Flexible Instruments

Recently, the use of robotic assisted

systems

Mintenbeck, Julien

for minimally invasive surgery has increased. This is due to the precise manipulation and miniaturisation of the used equipment. The systems can be divided into multi-port and single-port robots, both with the aim of trauma reduction but with different approaches. On the one hand there are multiple robots, one with each instrument or camera and on the other hand there

are

Ledermann, Christoph

small flexible instruments and only one incision in the abdominal wall. In this paper, a novel robotic system is presented that combines in a hybrid way the prior

described

Estaña, Ramon

characteristics and

continues

Heinz Wörn

the medical aims of trauma reduction and surgeon comfort. One single lightweight robot with three individual configurable flexible instruments is being researched. Additionally, the modular hardware and software architecture is based on the open-source idea and rapid-manufacturing processes.

Julien Mintenbeck, Christoph Ledermann, Ramon Estaña, Heinz Wörn

### Diagnosis Makes the Difference for a Successful Execution of High-Level Robot Control Programs

Faults in action execution and perception that occur at runtime negatively affects the probability that an agent is able to finish a

Mühlbacher, Clemens

successfully. There are several techniques such as hand-coded error recovery or diagnosis and repair approaches to deal with this problem. In this paper we present an experimental

comparison

Steinbauer, Gerald

of the fault tolerance of the popular (CRAM) framework using hand-coded error recovery and an approach using model-based diagnosis. The approaches were evaluated using a simulated robot delivery domain. Experimental results confirm that a control approach using diagnosis is able to significantly increase the success-rate and outperforms hand-coded error strategies.

Clemens Mühlbacher, Gerald Steinbauer

### Estimation of Needle Deflection in Layered Soft Tissue for Robotic Needle Steering

Precise needle tip placement is

important

Lee, Hyosang

in robotic needle steering but it is challenging to estimate the behavior of the flexible needle when it interacts with the soft tissue. Numerous studies have been focused on a needle

deflection

Kim, Jung

model in homogeneous tissue. However, human soft tissues generally have layered structure which has various mechanical properties and geometries. In this paper, we proposed a needle deflection model in a layered elastic medium considering variable needle–tissue interaction forces such as friction of needle surface, needle tip force, and lateral stiffness of soft tissue. For validation, needle steering experiment was performed to a layered tissue phantom and porcine meat which has a skin layer and muscle layer. Using the forces measured from the needle base in the experiment, the needle tip position was simulated. The simulated needle tip trajectory was then compared to the measured needle tip trajectory. The average needle tip estimation error was 1.78

$$\pm$$

±

0.84 mm for tissue phantom and 1.85

$$\pm$$

±

0.73 mm for porcine tissue. The proposed model could be applied to the robotic needle steering in layered tissue.

Hyosang Lee, Jung Kim

### Online Event Classification for Liver Needle Insertion Based on Force Patterns

In recent years percutaneous

treatments

Elgezua, Inko

for cancer have won momentum in the medical field. With it new needle insertion robots appeared to overcome the difficulties

associated

Song, Sangha

with needle

insertion

Kobayashi, Yo

into soft tissue. At first, the main

focus

Fujie, Masakatsu G.

was to achieve high needle placement accuracy, however, the focus nowadays has shifted toward needle steering and patient specific needle tissue interaction. In this paper we present a classification method to detect the type of tissue being punctured in real time. The purpose of the proposed method is to detect particular events that can be used in a situational awareness agent. First, we will introduce the methodology to create the statistical models used for classification, next, we prove the feasibility of the proposed classification method with experimental results and show that the proposed method hit a target even when tissue is deformed by analyzing needle insertion force patterns.

Inko Elgezua, Sangha Song, Yo Kobayashi, Masakatsu G. Fujie

### Anthropomorphic Finger Mechanism with a Nonelastic Branching Tendon

To realize both grasp

stability

Yanagisawa, Kazuya

and manipulation

dexterity

Ikemoto, Shuhei

is a central problem in the

development

Shirafuji, Shouhei

of robot hands. In recent years, many underactuated robot hands have

been

Hosoda, Koh

developed that flexibly conform to an object’s surface with simple control. In contrast, it is difficult to realize dexterous manipulation by such underactuated hands in which all degrees of freedom (DoFs) should be controlled. In this research, to realize the dexterous manipulation by simple mechanism and control, we develop a robot gripper comprising of two tendon-driven robotic fingers with nonelastic branching tendons. The branching tendon is a tendon that branches out and connects an actuator to different links. The two joints of this robotic finger are coupled by the nonelastic branching tendon when no external force is exerted. If sufficient external force is applied to the fingertip, one of the tendons slackens and the coupling between the two joints is lost. This means that the two-DoF robotic finger is easily controlled as a single DoF mechanism when reaching toward an object, but when the fingertip is placed on the object, the coupling provided by the branching tendon is released and the finger shifts. Based on this idea, we develop and control a two-DoF robotic finger equipoed with two tendons including a nonelastic branching tendon. We also analyze the conditions, where the branching tendon slacken, and confirmed in an experiment. As the result, the availability of controlling the slack of branched tendon was successfully confirmed.

Kazuya Yanagisawa, Shouhei Shirafuji, Shuhei Ikemoto, Koh Hosoda

### GMM-Based Single-Joint Angle Estimation Using EMG Signals

This paper aims to explore the possibility to use Electromyography (EMG) to train a

Gaussian

Michieletto, Stefano

Mixture Model (GMM) in order to estimate the bending angle of a single human joint.

In

Tonin, Luca

particular, EMG signals from eight leg muscles and the knee joint angle are acquired during a kick

Antonello, Mauro

from three different subjects. GMM is validated on new unseen data and the classification performances

are

Bortoletto, Roberto

compared with respect to the number of EMG channels and the number

of

Spolaor, Fabiola

collected trials used during the training phase. Achieved

results

Pagello, Enrico

show that our framework

is

Menegatti, Emanuele

able to obtain high performances even using few EMG channels and with a small training dataset (Normalized Mean Square Error: 0.96, 0.98, 0.98 for the three subjects, respectively), opening new and interesting perspectives for the hybrid control of humanoid robots and exoskeletons.

Stefano Michieletto, Luca Tonin, Mauro Antonello, Roberto Bortoletto, Fabiola Spolaor, Enrico Pagello, Emanuele Menegatti

### Human Muscle-Tendon Stiffness Estimation During Normal Gait Cycle Based on Gaussian Mixture Model

The aim of this study is to

estimate

Bortoletto, Roberto

the stiffness of the muscle-tendon unit, of human lower limb, during the execution of a normal gait cycle. Unlike the

analytical

Michieletto, Stefano

techniques already widely validated in literature and discussed below, a probabilistic

approach

Pagello, Enrico

based on the Gaussian Mixture Model (GMM) has been adopted here for the

computation

Piovesan, Davide

of the muscle-tendon unit stiffness. The obtained results for the major muscle groups are shown. The effectiveness of the proposed approach has been evaluated by computing the Root Mean Square (RMS) error between the stiffness calculated analytically and those calculated using the GMM, for each subject.

Roberto Bortoletto, Stefano Michieletto, Enrico Pagello, Davide Piovesan

### A Weak Generalized Inverse Applied to Redundancy Solving of Serial Chain Robots

A peculiar form of right

inverse

Tondu, Bertrand

derived from the theory of rectangular matrix determinants is considered instead of the classic Moore–Penrose pseudodinverse with the aim to get compact symbolic expressions for the redundancy solving of serial chain robots. Such an approach, based on the closed-form expressions of the

$$m\times m$$

m

×

m

minors of the

$$m\times n$$

m

×

n

robot Jacobian (

$$m< n)$$

m

<

n

)

and on its

$$(m-1)\times (n-1)$$

(

m

-

1

)

×

(

n

-

1

)

rectangular minors, is proposed as a possible new way for fast computation in inverse kinematic control.

Bertrand Tondu

### Interaction of Two Oscillator Aggregations

The main feature that keeps states and

structures

Yamauchi, Sho

stable can be seen in living organisms. This adjusting and adaptive features are called homeostasis. This integrated adaptive

feature

Kawamura, Hidenori

is achieved by the cooperation of organs in living organisms. Living organisms in nature

act

Suzuki, Keiji

dynamically due to this feature. Highly adaptive behavior caused by this feature is also observed in simple living organisms that have no neural circuits such as amoebas. Based on these facts, a method of control to generate homeostasis in robotic systems is proposed by assuming a robot system is an aggregation of oscillators and each parameter in a robot system is allocated to an oscillator. Especially, interaction between two independent robots as oscillator aggregation is focused in this paper.

Sho Yamauchi, Hidenori Kawamura, Keiji Suzuki

### Design and Control of a Vertical Ball Juggling Delta Robot Without Visual Guidance

In this paper, the design and control of a vertical

ball

Shareef, Zeeshan

juggling Delta robot is presented. The position and velocity of the ball, factors play an important role during juggling. Quite often these

factors

Just, Viktor

are calculated

using

Teichrieb, Heinrich

visual guidance. This paper introduces a control algorithm to juggle the

ball

Trächtler, Ansgar

vertically in two dimensions (2

D

) without visual guidance method. Instead of normal visual guidance method, an observer based on the reflection laws is used to get the continuous position and velocity of the ball. The next hitting time and the hitting velocity of the ball are predicted using the projectile motion equations. Three different controllers are designed for the stability and tracking of variable reference height of the ball during juggling and to keep the ball from falling off. The validation of this proposed control algorithm for ball juggling is shown by the simulation and preliminary experimental results.

Zeeshan Shareef, Viktor Just, Heinrich Teichrieb, Ansgar Trächtler

### On the Role of Compliance in Force Control

This paper proposes an overview and an

interpretation

Calanca, Andrea

on the role of compliance in force control within a framework where adaptive control arise as an intuitive approach. In our analysis, we show that force control

stability

Fiorini, Paolo

can be assured only if exists a compliant interface between the robot and the environment. Also, we prove that compliance is helpful to ensure well-defined force control dynamics, if combined with a low robot inertia. Otherwise adaptive control algorithms are proposed as a tool to deal with environment uncertainties. Finally, an experimental comparison between the adaptive approach and state of the art solutions is proposed.

Andrea Calanca, Paolo Fiorini

### iSplash-OPTIMIZE: Optimized Linear Carangiform Swimming Motion

This paper presents a new robotic fish,

iSplash

-OPTIMIZE, which is 0.6 m in

body

Clapham, Richard James

length and deploys a single actuator to drive discrete links across the full-body length. The main focus is on optimizing the

kinematics

Hu, Huosheng

parameters of its linear carangiform swimming motion in order to improve the distance travelled per beat. The experimental results show that the fish can be actuated at high frequencies up to 20 Hz due to deploying a continuous rotary power source. Each discrete link is able to be precisely tuned, providing accurate kinematics with little mechanical loss.

Richard James Clapham, Huosheng Hu

### Control of a Four-Forked Steering Walker—Design of Virtual Mechanical Elements Based on Desired Motions

This paper presents a new undulatory locomotor, a four-forked steering walker, and introduces a quite new design methodology of virtual mechanical elements which facilitates to achieve desired motions. The four-forked steering

walker

Yamaguchi, Hiroaki

transforms periodic changes in shape into movement as if it performs roller skating. The virtual

mechanical

Takahashi, Ryosuke

elements are originally defined solely for conversion of kinematical equations of the locomotor into a chained form. It is shown in this paper that such definition has another aspect in which the virtual mechanical elements

enabling

Kawakami, Atsushi

the conversion can be also designed according to desired motions. Specifically, the base of the locomotor can be replaced with a virtual base and its shape and locations of the virtual mechanical elements on the virtual base can be modified according to path-following and gyrating motions. The validity of the design methodology is verified experimentally.

Hiroaki Yamaguchi, Ryosuke Takahashi, Atsushi Kawakami

### Collision Avoidance with Task Constraints and Kinematic Limitations for Dual Arm Robots

Human-robot interaction (HRI) is a key

element

Ceriani, Nicola Maria

for diffusion of robotised production. Clear advantages in

flexibility

Zanchettin, Andrea Maria

and productivity are possible, when the two operators are free to interact, as they are endowed with complementary skills. To achieve such a goal, safety

systems

Rocco, Paolo

capable of coping with task and robot constraints have to be designed. In this paper, a collision avoidance strategy, tackling consistency with task constraints and robot kinematic limitations, is proposed. Robot joint velocities are selected with a QP optimisation problem, minimising the difference from evasive velocities, while respecting task constraints. Integration with an industrial controller is discussed as well, while the strategy is experimentally validated on a dual arm industrial robot prototype, working in close interaction with a human.

Nicola Maria Ceriani, Andrea Maria Zanchettin, Paolo Rocco

### Reactive Constrained Model Predictive Control for Redundant Mobile Manipulators

Research interest in redundant mobile manipulators has been constantly increasing during the last decade. The opportunities offered by the redundant degrees of freedom, together with the exploitation of the mobile base, would allow such robots to complete their main task while complying with additional tasks or constraints. These features would make it easier for robots to work in a partly unstructured and dynamic environment, thus increasing production flexibility. In this work, a reactive constraint-based control strategy for mobile manipulators is proposed, which accomplishes a positioning task while simultaneously avoiding unknown and unpredictable obstacles. Differently from other approaches, the trajectory is computed exclusively online, by exploiting the MPC method, without the need of a pre-planned path. Experimental verification on a KUKA youBot shows the applicability of the approach.

Giovanni Buizza Avanzini, Andrea Maria Zanchettin, Paolo Rocco

### Analysis and Simulation of a Jumping Robot Actuated by Shape Memory Alloy

This paper reports design, analysis, and

simulation

Ho, Thanhtam

of a mesoscale robot that is designed to locomote by jumping. Jumping can be quite advantageous over other locomotion ways on the ground, especially in terms of ability of

obstacle

Lee, Sangyoon

avoidance. First bioinspired design of a jumping mechanism using only one shape memory alloy (SMA) spring is introduced. The use of SMA spring reduces the weight and the complexity of the robot. Then design of a landing mechanism and analysis on stable landing are described. The robot is designed in order for its center of gravity (COG) to lie inside the safety cone. Methods for choosing optimum parameters in the robot design such as the spring stiffness and the tilt angle are also presented through mechanical analysis and simulation.

Thanhtam Ho, Sangyoon Lee

### Parametric Continuous Curvature Path for Smooth Steering with Car-like Vehicles

This paper proposes a solution to obtain parametric continuous curvature path for nonholonomic car-like vehicle and addresses also efficient algorithm to obtain an appropriate reference path for smooth path following. As a local planner using parametrically adjustable clothoid, the proposed solution, of constant sharpness, permits to obtain continuous curvature and thus smooth steering behavior of the vehicle which enhances the passenger comfort. The proposed parametric continuous curvature path is integrated in a global planner which includes an appropriate way of isolating the boundary conditions of each local planner. This last characteristic is useful to reconstruct a smooth reference path from raw data of actual vehicle trajectory. The solution is applied to a real road datasets and tested using

shape Lyapunov

based controller for validating the effectiveness on path following performance. The simulated results show reliable and enhanced performance for vehicle path following when the proposed algorithms are used.

Suhyeon Gim, Lounis Adouane, Sukhan Lee, Jean-Pierre Derutin

### Diversive Curiosity in Robots and Action Selection Method for Obtaining Unexperienced Sensory Information

Humans can acquire new knowledge by themselves, and robots are also expected to have such ability by

introducing

Morisawa, Toma

curiosity. In

previous

Mizuuchi, Ikuo

researches, curiosity is expressed as just a judging system that decides whether the robot continues looking over a specific environment. In this paper we define curiosity as a direction in the multidimensional sensory space, and propose an action selection method based on the curiosity vector. The system estimates the relation between action and resulting sensory changes using the stored observed data. It selects an action so as to obtain the desired sensory information, based on the curiosity vector. This paper also describes experiments in which a wheeled mobile robot moves toward an unknown area and a humanoid discovers actions to obtain unexperienced sensory information.

Toma Morisawa, Ikuo Mizuuchi

### Cognition-Enabled Robot Control for Mixed Human-Robot Rescue Teams

In this paper we present a cognition-enabled control framework for robot control in mixed human-robot teams performing rescue missions after avalanches. We could

focus

Yazdani, Fereshta

on two key reasoning mechanisms: First, reasoning about the robot capabilities, which allow them to make best use of the

hardware

Brieber, Benjamin

and software components they are equipped with. Second, context-directed interpretation of vague commands, which enables the human leader of the rescue team to state tasks naturally. Simulation-based reasoning mechanisms

then

Beetz, Michael

refine the vague and ambiguous instructions in the given capability context to appropriate task interpretations. We could show that by employing these reasoning mechanisms we can specify generic plans that automatically adapt themselves to the robotic agent executing them.

Fereshta Yazdani, Benjamin Brieber, Michael Beetz

### Autonomous Construction with Compliant Building Material

In this paper, we develop an

autonomous

Soleymani, Touraj

construction system in which a self-contained ground robot builds a protective

barrier

Trianni, Vito

by means of compliant pockets (i.e., filled bags). We present a stochastic control algorithm based on two biological mechanisms (stigmergy and templates) that

takes

Bonani, Michael

advantage of compliant pockets for autonomous construction. The control algorithm guides the robot to build the structure without relying on any external motion capture

system

or external computer. We propose a statistical model to represent the structures built with the compliant pockets, and we provide a set of criteria for assessing the performance of the proposed system. To demonstrate the

feasibility

Dorigo, Marco

of the proposed system, real-robot experiments were carried out. In each experiment, the robot successfully built the structure. The results show the viability of the proposed autonomous construction system.

Touraj Soleymani, Vito Trianni, Michael Bonani, Francesco Mondada, Marco Dorigo

### Performing Assembly Task Under Constraints Using 3D Sensor-Based Control

Basic robotic operations such as grasps of structure’s

parts

Vandernotte, Sylvain

and their placement are mostly model based and do not take into account

modelling

Chriette, Abdelhamid

error and geometry variations. As consequences, quality of the

assembly

cannot be trusted. In this paper, we propose to use 3D sensor and sensor-based

control

Martinet, Philippe

to improve the precision a positioning task. The interaction matrix is constructed upon assembly constraint definition. Those assembly constraints are taken into a task sequencing mechanism, with others robot-specific and environment tasks. The positioning operation is tested with Gazebo simulator and ROS environment and demonstrate the effectiveness of the approach.

Sylvain Vandernotte, Abdelhamid Chriette, Adolfo Suarez Roos, Philippe Martinet

### From Stigmergy to Affordance: The Mechanical Basis of Robot Motion Control

In the last decade, the development of multi robot systems has shown with growing evidence how a well-balanced deliberative-reactive coordination can provide a group of

robots

D’Angelo, Antonio

with an efficient and robust collective behavior. In this paper, we want to cover exhaustively this issue from the point of view of the single agent in the general model we have

Pagello, Enrico

presented as “roboticle framework”. Starting from a pure mechanical interpretation of an autonomous robot motion we shall understand the notions of stigmergy and affordance by maintaining at the sub-symbolic level all the relevant information useful to drive properly the robot while it participates to the collective action. Specifically, we shall focus on some interesting parameters through which the designer of the single robot governor’s unit could be helped to trigger its individual behavior within a collective scenario.

Antonio D’Angelo, Enrico Pagello

### Teleoperation of On-Road Vehicles via Immersive Telepresence Using Off-the-shelf Components

The quality of visual information and

response

Shen, Xiaotong

time are crucial aspects of any modern teleoperation system. This is especially

true

Chong, Zhuang Jie

for operation of on-road vehicles, which must function in highly dynamic,

unforgiving

Pendleton, Scott

environments. In this

work

Frazzoli, Emilio

we

demonstrate

Fu, Guo Ming James

that a suitable teleoperation system can

be

Qin, Baoxing

exclusively composed of low-cost off-the-shelf components yet still meet the high performance

demands

Ang Jr., Marcelo H.

of remotely driving a car on the road. The user is given immersive situational awareness through an on-board head-mounted display linked to an actuated stereoscopic camera, thereby maintaining depth perception and intuitive camera control. Communication speeds are evaluated over various wireless connection types, and a usability study shows that the system allows for advanced driving maneuvers while remotely controlled. 3G and 4G data networks are demonstrated to provide adequate bandwidth for the task given proper data compression, thus expanding the potential range for teleoperation. Applications for such a system are further discussed, extending to fleet management and autonomous vehicle safety measures.

Xiaotong Shen, Zhuang Jie Chong, Scott Pendleton, Guo Ming James Fu, Baoxing Qin, Emilio Frazzoli, Marcelo H. Ang

### Network Controller for Teleoperated Mobile Robotic Agents

In the communication channels of bilateral teleoperation

systems

Márton, Lőrinc

it is critical to keep the delay and delay

variation

Haller, Piroska

under prescribed limits. Otherwise, the

stability

Szabó, Tamás

and

transparency

Sándor, Hunor

of the teleoperation are compromised. In this

paper

Vajda, Tamás

a network controller

is

Szántó, Zoltán

proposed for video supported teleoperation systems implemented over WLANs (Wireless Local Area Networks). The algorithm controls the transfer rates in the video channels which serve for online monitoring of the teleoperation task. The video transfer rate controller was developed using multi-objective optimization and it assures the best achievable video transmission quality for a prescribed delay or jitter in the communication channels of the teleoperator. The performed real-time experimental measurements show that, with the proposed rate control algorithm, the delay and its fluctuation can be substantially reduced in the teleoperation systems implemented over WLAN.

Lőrinc Márton, Piroska Haller, Tamás Szabó, Hunor Sándor, Tamás Vajda, Zoltán Szántó

### Development of Mobile Robots Using Off-the-Shelf Open-Source Hardware and Software Components for Motion and Pose Tracking

Robot development could benefit from reuse of hardware and

software

Cucci, Davide A.

components. One of the most successful example is the Robot Operating System (ROS), which aims at supporting robot development and

research

Migliavacca, Martino

in robotics by

providing

Bonarini, Andrea

a modular and

general

Matteucci, Matteo

purpose software framework; yet, at a lower level, i.e., mechanics, electronics, and firmware, component-based robotics is still in its early stages. Moreover, basic functionalities, such as robust odometry and pose tracking, seem to be solved on an ad-hoc basis, while no generic solution can be plugged in a system as an off-the-shelf component. This paper presents two modular tools for the development of robotics applications and how these can be deployed on different robot platforms as off-the-shelf, open hardware and open software, components at the hardware, firmware, and core software levels. A case study based on two mobile platforms is described showing how it is possible to reuse physical components in the building of a mobile robot and provide an off-the-shelf architecture which sports multi-sensor fusion for robust odometry and pose tracking, requiring a fraction of the effort with respect to ad hoc techniques.

Davide A. Cucci, Martino Migliavacca, Andrea Bonarini, Matteo Matteucci

### Whole-Body Robot Motion Learning by Kinesthetic Teaching

Learning whole-body robot

motion

Lin, Hsien-I

is a challenging task because balance control should be taken into consideration. An

intuitive

Chen, Yung-Yao

way to teach motion to a humanoid robot is to apply human demonstration data to the robot. Since balance control was usually done by presetting the zero-moment-point (ZMP) trajectory of a robot, the challenge became the

conversion

Huang, Yu-Che

problem from human motion to robot motion, making the ZMP trajectory satisfy the stability. In this paper, we use kinesthetic teaching to teach whole-body robot motion by directly pulling the limbs of a robot without any conversion from human to robot motion. To keep the robot balanced, we propose a trade-off function by considering motion similarity and balance simultaneously and adopt the genetic algorithm (GA) to find the solution for adapting the taught motion. We validated the proposed method on an Aldebaran NAO robot and the results showed that the robot was taught to perform side and back kicks via kinesthetic teaching.

Hsien-I Lin, Yung-Yao Chen, Yu-Che Huang

### Experiment of Stereo Sensors for Chemical Plume Tracing by Optogenetic Silkworm Moth

In this paper, we introduced an experimental

system

Kishi, Kotaro

in which an optogenetic moth was bounded to a virtual environment that we could clearly control and observe. By using this system, we

investigated

Kurabayashi, Daisuke

the effect of

stereo

Minegishi, Ryo

sensors for the

performance

Sakurai, Takeshi

of chemical

plume

Kanzaki, Ryohei

examined

Tabuchi, Masashi

the three

different

Sezutsu, Hideki

sensory conditions,

Normal

,

Reverse

, where the left and the right sensors were inverted, and

Both

, where a moth always received both the left and the right sensor inputs simultaneously. The result showed us that (i) the moth’s CPT behavior was highly fault-tolerant against the jamming to the stereo sensors, and (ii) consistent behavior might be important for the CPT performance.

Kotaro Kishi, Daisuke Kurabayashi, Ryo Minegishi, Takeshi Sakurai, Ryohei Kanzaki, Masashi Tabuchi, Hideki Sezutsu

### Distributed Algorithm for Robotic Network Self-deployment in Indoor Environments Using Wireless Signal Strength

Mobile robots with wireless

capabilities

Miyagusuku, Renato

can enable network connectivity over large areas by retransmitting wireless

signals

Yamashita, Atsushi

from a ground station. Our goal is, for such robotic networks, to enhance current teleoperated robots’ ability to perform reconnaissance or assist

human

Asama, Hajime

first responders on victims’ search and rescue operations. On these missions, uninterrupted communications between teleoperated robots and their human operators are essential. In order to maximize the teleoperated robots’ working area, the robotic network has to deploy itself, in an unknown and potentially hazardous environment, spreading as much as possible without losing network connectivity. In this paper, we consider an urban search and rescue setting; and present a distributed algorithm that allows simple mobile robots to self-deploy and create robotic networks, without the need of advanced self-localization capabilities nor prior knowledge of the environment.

Renato Miyagusuku, Atsushi Yamashita, Hajime Asama

### Adaptive Design and Control of a Robot-Assisted Lower Back Exoskeletal Spine System

In order to help elder people who

suffer

He, Fuben

from lower back pain caused by lower spine degeneration, a novel kind of robot-assisted exoskeleton spine was designed. It was mainly applied to lift their upper bodies for

assisting

Zhang, Haohan

movements and reducing backache

during

Bortoletto, Roberto

walking. The aim of this system was to control an elastically actuated

motor

Liang, Yande

to provide extra torques on a user’s hip by following the gaits in locomotion. And the whole exoskeletal spine mechanism (exo-spine) has been

built

Pagello, Enrico

of flexible material and fixed on an artificial pelvis. Thanks to the use of a cable-pulley-spring structure the torque applied to the hip is greatly amplified and would eventually affect the deformation of exo-spine, so that an auxiliary force is generated on the lower back to support user’s spine during the movements. Although the overall robot-assisted system was easily imaged and designed, its intrinsic complexity needed careful analysis, because the actuating process becomes highly nonlinear and noisy when compliant movements are demanded to mimic human performances in locomotion. Therefore, some appropriate assumptions were introduced, and to enhance the robustness of system, an adaptive controller was designed by applying Lyapunov Stability Theory. Finally, the correctness and feasibility of our proposed system were tested and estimated through a set of experimental simulations.

Fuben He, Haohan Zhang, Roberto Bortoletto, Yande Liang, Enrico Pagello

### Toward Autonomous Robots for Demolitions in Unstructured Environments

The construction industry is a capital-intensive sector that is steadily turning toward mechanized and automated solutions in the past few decades. However, due to some specificities of this field, it is still technologically behind other sectors, such as manufacturing. Robotic technologies provide room for improvements, that could lead to economical, technical, and also social benefits. We present a possible

conceptual

Corucci, Francesco

framework for an

autonomous

Ruffaldi, Emanuele

robot for indoor demolitions, featuring enhanced perceptual capabilities, situational awareness, as well as intuitive Human–Robot Interaction (HRI) paradigms. The paper deals with several aspects of the demolition task, ranging from perception, to planning, to HRI. With respect to perception, we focus on the design and development of some of the perceptual capabilities needed for such robots, as essential to support autonomy, safety, and situational awareness in unstructured construction sites. Particularly, we propose a novel segmentation algorithm that allows the robot to work in highly unstructured scenarios, as well as a mechanism for detecting and quantifying spatial changes during the task. As far as HRI is concerned, a novel interaction paradigm based on laser designation is proposed. Proposed concepts were implemented and tested on a real, scaled-down, controlled mock-up that, while simplifying some aspects of the task, is able to mimic some general characteristics of a real demolition scenario. From lessons learned in this controlled environment we point out some requirements and foreseen issues in facing the complexity of a real demolition setup.

Francesco Corucci, Emanuele Ruffaldi

### From Insects to Micro Air Vehicles—A Comparison of Reactive Plume Tracking Strategies

Insect behavior is a common source of

inspiration

Neumann, Patrick P.

for roboticists and computer scientists when designing gas-sensitive mobile robots. More specifically, tracking

airborne

Bennetts, Victor Hernandez

odor plumes, and localization of distant gas

sources

Lilienthal, Achim J.

are

abilities

Bartholmai, Matthias

that suit practical applications such as leak localization and emission monitoring. Gas sensing with mobile robots has been mostly addressed with ground-based platforms and under simplified conditions and thus, there exist a significant gap between the outstanding insect abilities and state-of-the-art robotics systems. As a step toward practical applications, we evaluated the performance of three biologically inspired plume tracking algorithms. The evaluation is carried out not only with computer simulations, but also with real-world experiments in which, a quadrocopter-based micro Unmanned Aerial Vehicle autonomously follows a methane trail toward the emitting source. Compared to ground robots, micro UAVs bring several advantages such as their superior steering capabilities and fewer mobility restrictions in complex terrains. The experimental evaluation shows that, under certain environmental conditions, insect like behavior in gas-sensitive UAVs is feasible in real-world environments.

Patrick P. Neumann, Victor Hernandez Bennetts, Achim J. Lilienthal, Matthias Bartholmai

### A Proposed Architecture for Autonomous Operations in Backhoe Machines

In this work is developed an

architecture

Mastalli, Carlos

which consists of four main components: perception system, tasks planning, motion planner, and control systems that allow autonomous operations in backhoe machines. In the first part is described the architecture of control system. A set of techniques for collision mapping of the scene is described and implemented. Moreover, a motion planning system based on Learning from Demonstration using Dynamic Movement Primitives as control policy is proposed, which allows backhoe machines to perform operations in autonomous manner.

A

Fernández-López, Gerardo

statement of reasons is presented, wherein we justified the implementation of such motion system versus planners like

$$\text {A}^*$$

A

, Probabilistic RoadMap (PRM), Rapidly-exploring Random Tree (RRT), etc. In addition, we present the performance of the architecture in a simulation environment.

Carlos Mastalli, Gerardo Fernández-López

This paper presents the

application

Mastrogiovanni, Fulvio

of a minimalistic navigation strategy, based on the well-known BUG2 algorithm, to solve the

problem

Sgorbissa, Antonio

of reaching a goal

position

Zaccaria, Renato

in a multi-floor indoor scenario using a quadrotor. Examples of this scenario include buildings and in general

cluttered

Marino, Roberto

indoor areas. As far as energy backup is concerned the quadrotor shows stricts constraints: for this reason implementing a low-consumption navigation strategy is a major issue. We present a two-layer navigation strategy, called MF-BUG2, useful to navigate in multi-floor buildings starting from the ground floor toward the last or vice versa while searching for an interesting physical quantity (i.e,. gas leak, electromagnetic source). In the lower layer a BUG-like algorithm is able to drive the flying robot, equipped with a salient-cue sensor and a laser-range-finder, toward the estimated position of goal on the horizontal plane while avoiding obstacles and using minimal computational power and memory (the boundary-following behavior uses an Artificial Potential Field to navigate around the obstacles). If the estimated goal position is reached but the salient-cue-sensor does not detect a salient quantity the higher level of the planner calls Dijkstra algorithm to computes the minimum-distance path to change the floor, assuming to know in advance the 2D position of the passages among different floors, and then moves vertically. The overall strategy is usefull for

indoor inspection

in hazardous scenarios. The algorithm is validated in simulation, investigating the robustness with respect to the laser-range-finder noise.

Roberto Marino, Fulvio Mastrogiovanni, Antonio Sgorbissa, Renato Zaccaria

### Formal Verification of Behaviour Networks Including Hardware Failures

The paper deals with the problem of

verifying

Kiekbusch, Lisa

a behaviour-based

control

Armbrust, Christopher

system.

Although

Berns, Karsten

failures in sensor hardware and software can have strong influences on the robot reaction, they are often neglected in the verification process. Instead, perfect sensing is assumed. Therefore, this paper provides an approach for modelling the sensor chain in a formal way and connect it to the formal model of the control system. This model can be verified using model checking techniques, which is shown on the example of the control system of the autonomous off-road robot

ravon

(

ravon

:

R

obust

A

utonomous

V

ehicle for

O

N

avigation).

Lisa Kiekbusch, Christopher Armbrust, Karsten Berns

### Mapping with Micro Aerial Vehicles by Registration of Sparse 3D Laser Scans

Micro aerial vehicles (MAVs) pose specific constraints on onboard sensing, mainly limited payload and limited processing power. For accurate 3D

mapping

Holz, Dirk

even in GPS denied environments, we have designed a light-weight 3D laser scanner specifically for the application on MAVs. Similar to other custom-built 3D

laser

Behnke, Sven

scanners composed of a rotating 2D laser range finder, it exhibits different point densities within and between individual scan lines. When rotated fast, such nonuniform point densities influence neighborhood searches which in turn may negatively affect local feature estimation and scan registration. We present a complete pipeline for 3D mapping including pairwise registration and global alignment of 3D scans acquired in-flight. For registration, we extend a state-of-the-art registration algorithm to include topological information from approximate surface reconstructions. For global alignment, we use a graph-based approach making use of the same error metric and iteratively refine the complete vehicle trajectory. In experiments, we show that our approach can compensate for the effects caused by different point densities up to very low angular resolutions and that we can build accurate and consistent 3D maps in-flight with a micro aerial vehicle.

Dirk Holz, Sven Behnke

### Learning Manipulation by Sequencing Motor Primitives with a Two-Armed Robot

Learning to perform complex tasks out of a

sequence

Lioutikov, Rudolf

of simple small demonstrations is a key ability for more flexible robots. In this paper, we present a system that allows for

the

Kroemer, Oliver

acquisition of such task executions based on dynamical movement primitives (DMPs). DMPs are a successful approach to encode and generalize robot movements. However, current applications involving DMPs

mainly

Maeda, Guilherme

explore movements that, although challenging in terms of dexterity and dimensionality, usually

comprise

Peters, Jan

a single continuous movement. This article describes the implementation of a novel system that allows sequencing of simple demonstrations, each one encoded by its own DMP, to achieve a bimanual manipulation task that is too complex to be demonstrated with a single teaching action. As the experimental results show, the resulting system can successfully accomplish a sequenced task of grasping, placing and cutting a vegetable using a setup of a bimanual robot.

Rudolf Lioutikov, Oliver Kroemer, Guilherme Maeda, Jan Peters

Very few times a short

introduction

like this can tell a user so many things about a project. In this paper we introduce “Alleine” (“alone” in English), an autonomous wheelchair that can be voice- and brain-driven,

capable

Rojas, Raúl

of navigating in unstructured environments, avoid unexpected obstacles, and reach a desired goal like “kitchen” or “living room.” Our prototype is the very first to participate in a domestic service robots category as robocup@home 2013, by using a small notebook laptop as the main processor.

### Kinematic Analysis of a 3D Printable 4-DOF Desktop Robot Actuated Exclusively by Revolute Pairs

This paper describes the kinematic

structure

DallaLibera, Fabio

of Yondy, a

desktop

Penaloza, Christian I.

robot with three

rotational

Yoshikawa, Yuichiro

DOFs and

one

Ishiguro, Hiroshi

translational DOF. The kinematic structure comprises only rotational joints, easing its construction using off-the-shelf rotational actuators like servomotors. No pinion-rack or other mechanical elements are required for its construction, permitting its realization even with low-accuracy 3D printers. The robot is realized as a hybrid 4-DOF mechanism, with a planar 2-DOF parallel manipulator connected in series with other 2 DOFs. Particular focus is given to the parallel manipulator, which is interesting from a theoretical point of view because it is able to undergo nonsingular assembly mode transitions.

Fabio DallaLibera, Christian I. Penaloza, Yuichiro Yoshikawa, Hiroshi Ishiguro

### Dynamic Simulation of Robotic Devices Using the Biomechanical Simulator OpenSim

The efforts in robotic devices simulation are constantly increasing in the latest years. A

reliable

Vivian, Michele

and accurate dynamic simulator could improve the quality of the mechanical and

controller

Tagliapietra, Luca

design, while reducing the cost of development. Moreover, this tool should be independent from the specific robotic device

and

Sartori, Massimo

quite flexible and advanced to support the recent research on complex biologically inspired systems. Based on

these

Reggiani, Monica

considerations we evaluated the possibility to use OpenSim, a simulation tool for the biomechanical community, to implement robot dynamic simulations. OpenSim is a free software that provides a platform on which the biomechanics community can build and share libraries of human movement dynamic simulation. Currently, OpenSim is still not largely used in robotic. Nevertheless, it would have several advantages for the large libraries of tools and model of biological bodies. The tools to simulate the human movement could be used also in the simulation of the new generation of robots with variable stiffness and impedance actuators. So this work reports our first examples of robotic devices dynamic simulation using OpenSim. The obtained results and the interest of the OpenSim developers on our work encourage to follow this research direction.

Michele Vivian, Luca Tagliapietra, Massimo Sartori, Monica Reggiani

### RGB-D Human Detection and Tracking for Industrial Environments

Reliably detecting and tracking

movements

Munaro, Matteo

of nearby workers on the factory floor are crucial to the safety of advanced manufacturing automation in which humans and robots share the same workspace. In this work, we address the problem of multiple

people

Lewis, Christopher

detection and tracking in industrial environments by proposing algorithms which exploit

both

Chambers, David

color and depth data to robustly

track

Hvass, Paul

people in real time. For

people

Menegatti, Emanuele

detection, a cascade organization of these algorithms is proposed, while tracking is performed based on a particle filter which can interpolate sparse detection results by exploiting color histograms of people. Tracking results of different combinations of the proposed methods are evaluated on a novel dataset collected with a consumer RGB-D sensor in an industrial-like environment. Our techniques obtain good tracking performances even in an industrial setting and reach more than 30 Hz update rate. All these algorithms have been released as open source as part of the ROS-Industrial project.

Matteo Munaro, Christopher Lewis, David Chambers, Paul Hvass, Emanuele Menegatti

### Architecture Challenges for Intelligent Autonomous Machines

An Industrial Perspective

Machines are displaying a

trend

Behere, Sagar

of increasing autonomy. This has a far reaching impact on the architectures of the

embedded

Asplund, Fredrik

systems within the machine. The impact needs to be clearly understood and the main obstacles to

autonomy

Söderberg, Andreas

need to be identified. The obstacles,

especially

Törngren, Martin

from an industrial perspective, are not just technological but also relate to system aspects like certification, development processes and product safety. In this paper, we identify and discuss some of the main obstacles to autonomy from the viewpoint of technical specialists working on advanced industrial product development. The identified obstacles cover topics like world modeling, user interaction, complexity and system safety.

Sagar Behere, Fredrik Asplund, Andreas Söderberg, Martin Törngren

### Engineering Approaches and Methods to Verify Software in Autonomous Systems

We present three computer-augmented software engineering approaches to ensure dependability at

different

Cicala, G.

levels of control architectures in autonomous robots. For each approach, we outline the

methodological

Khalili, A.

framework, our

current

Metta, G.

achievements, and open issues. Albeit our results are still preliminary, we believe that

furthering

Tacchella, A.

research

along

Natale, L.

these lines can provide cost-effective

techniques

Pathak, S.

to make autonomous robots

safe

Pulina, L.

and thus fit for commercial purposes.

G. Cicala, A. Khalili, G. Metta, L. Natale, S. Pathak, L. Pulina, A. Tacchella

### A Knowledge-Based Approach to Crack Detection in Thermographic Images

This paper describes an innovative visual

inspection

Ghidoni, Stefano

system for the detection of small cracks in metal parts. Given the extremely low dimension of the defects to be detected, the

system

Antonello, Mauro

is based on a thermographic approach: defects are recognized analyzing the heat

flux

Nanni, Loris

induced by an excitation. The system is able to analyze parts of very high complexity, like a crankshaft, thanks to the introduction of an

articulated

Menegatti, Emanuele

robot, used for moving the part. The system also benefits from a deep knowledge of the inspected part and of the imaging system: this is exploited to reduce the high number of artifacts and reflections that appear in thermographic images when heat sources are employed. The core of the inspection mechanism is a computer vision algorithm that is capable of analyzing the thermographic images, extract the thermal information, and exploit a Support Vector Machine (SVM) classifier to provide a final decision on the presence of a crack in the analyzed part.

Stefano Ghidoni, Mauro Antonello, Loris Nanni, Emanuele Menegatti

### Backmatter

Weitere Informationen