Skip to main content

Über dieses Buch

This book constitutes the refereed proceedings of the 12th Latin American Robotics Symposium and Third Brazilian Symposium on Robotics, LARS 2015 / SBR 2015, held in Uberlândia, Brazil, in October/November 2015. The 17 revised full papers presented were carefully reviewed and selected from 80 submissions. The selected papers present a complete and solid reference of the state-of-the-art of intelligent robotics and automation research, covering the following areas: autonomous mobile robots, tele-operated and telepresence robots, human-robot interaction, trajectory control for mobile robots, autonomous vehicles, service-oriented robotic systems, semantic mapping, environment mapping, visual odometry, applications of RGB-D sensors, humanoid and biped robots, Robocup soccer robots, robot control, path planning, multiple vehicles and teams of robots.



Evaluating the Performance of Two Computer Vision Techniques for a Mobile Humanoid Agent Acting at RoboCup KidSized Soccer League

A humanoid robot capable of playing soccer needs to identify several objects in the soccer field in order to play soccer. The robot has to be able to recognize the ball, teammates and opponents, inferring information such as their distance and estimated location. In order to achieve this key requisite, this paper analyzes two descriptor algorithms, HAAR and HOG, so that one of them can be used for recognizing humanoid robots with less false positives alarms and with best frame per second rate. They were used with their respective classical classifiers, AdaBoost and SVM. As many different robots are available in RoboCup domain, the descriptor needs to describe features in a way that they can be distinguished from the background at the same time the classification has to have a good generalization capability. Although some limitations appeared in tests, the results were beyond expectations. Given the results, the chosen descriptor should be able to identify a mainly white-ball, which is clearly a simpler object. The results for ball detection were also quite interesting.
Claudio O Vilão, Vinicius Nicassio Ferreira, Luiz Antonio Celiberto, Reinaldo A. C. Bianchi

Dense Tracking with Range Cameras Using Key Frames

We present a low cost localization system that exploits dense image information to continuously track the position of a range camera in 6DOF. This work has two main contributions: First, the localization of the camera is performed with respect to a set of keyframes selected according to a spatial criteria producing a less populated and more uniform distribution of keyframes in space. This allows us to avoid the computational overload caused by having to estimate a depthmap at the frame rate of the camera as it is common in other dense sequential methods. Second, we propose a two-stage approach to compute the current location of the camera with respect to its closest keyframe. During the first stage, our system calculates an initial relative pose estimate from a sparse set of 3D to 2D point correspondences. This estimate is then refined during the second stage using a dense image alignment. The refinement step is stated as a Non Linear Least Squares (NLQs) optimisation embedded in a coarse to fine approach that minimizes the photo-consistency error between the current image and a warped version of the image associated to the closest keyframe and its depth map.
To validate the accuracy of our system, we conducted experiments using datasets with perfectly known trajectory and with both, perfect ray-traced images and images with noise and blur. We also evaluate the accuracy of the system using datasets with RGBD images taken at different frame-rates, and the improvements in convergence due to our coarse-to-find approach. Our assessment shows that our system is able to achieve millimeter accuracy. Most of the expensive calculations are carried out by exploiting parallel computation on a GPU.
Andrés Díaz, Lina Paz, Eduardo Caicedo, Pedro Piniés

Safe Navigation of Mobile Robots Using a Hybrid Navigation Framework with a Fuzzy Logic Decision Process

Autonomous navigation in dynamic environments is one of the most important problems in robotics. The different solutions to achieve this goal may be categorized into two big groups, deliberative methods and reactive methods. Deliberative methods require precise map knowledge, are computationally intensive, but they usually assure a path to the goal, on the other hand reactive methods are fast, dynamic but also subject to local minima among other problems. In this paper we propose a hybrid reactive-deliberative framework for mobile robots navigation which integrates the advantages of a high level deliberative planner with a reactive low-level control. The reactive layer of this new system uses the new map information in an asynchronous way allowing a much more dynamic response of the system to environment changes. For the merging of the reactive and deliberative behaviours a new Fuzzy Logic layer is proposed which defines the contribution of each navigation layer into the final movement of the robotic platform in real-time. The proposed framework was tested in a simulated Amigobot robot with a simulated Kinect sensor using the robotic simulation platform V-REP and the programming of the different layers was implemented in ROS.
Elvis Ruiz, Raul Acuña

A Harris Corner Detector Implementation in SoC-FPGA for Visual SLAM

The present paper discusses the implementation of the Harris and Stephen corner detector algorithm optimized for an embedded system-on-a-chip (SOC) platform that integrates a multicore ARM processor and FPGA fabric in a single chip, the Xilinx Zynq-7000. The algorithm is implemented as a hardware co-processor on the FPGA portion of the SoC. As a whole, the SoC is used as a stereo vision pre-processing module to retrieve depth information from the features in order to compose 3D landmark points for Visual SLAM, speeding up feature extraction and relieving this highly parallelizable process from the main embedded processor. The optimizations of the algorithm’s hardware implementation take into account the particularities of the SoC, such as compliance with its I/O requirements and FPGA’s constraints on the amount of logical elements available for hardware synthesis. Also, optimizations done in order to reduce the time of execution of the algorithm in hardware, such as parallelization and introduction of a pipeline, are also presented in the article. A speedup of 1.77 was achieved when comparing the time of execution of the algorithm in the hardware coprocessor with the algorithm running in software in the dual-core ARM processor.
Victor Hugo Schulz, Felipe Gustavo Bombardelli, Eduardo Todt

Comparison Among Experimental PID Auto Tuning Methods for a Self-balancing Robot

A self-balancing robot, also known as two-wheeled vehicle is an unstable system and it can be approximated to inverted pendulum, so there is a need of a suitable controller so that it can be stabilized. This paper compares five PID design techniques without mathematical model of the system in order to remain it stand. The PID tuning methods discussed are Manual, Ziegler-Nichols, Relay, Augmented Ziegler-Nichols and Augmented Relay. The augmented method modifies the PID constants online depending on the error value and use a Ziegler-Nichols or Relay PID tuned controller as initial one. Some experimental results presented suggest that the Ziegler-Nichols tuning method is slightly better than the other techniques. All the electronic gadgets and algorithms are embedded in the prototype.
Marcus Romano Salles Bernardes de Souza, Rodrigo Hiroshi Murofushi, José Jean-Paul Zanlucchi de Souza Tavares, José Francisco Ribeiro

RoSoS - A Free and Open-Source Robot Soccer Simulator for Educational Robotics

The use of robots as educational tools provides a stimulating environment for students. Some robotics competitions focus on primary and secondary school aged children, and serve as motivation for students to get involved in educational robotics activities. Although very appealing, many students cannot participate on robotics competitions because they cannot afford robotics kits. Hence, several students have no access to educational robotics, especially on developing countries. To minimize this problem and contribute to education equality, we have created RoSoS Robot Soccer Simulator, in which students program virtual robots in a similar way that they would program their real ones. In this chapter we explain some technical details of RoSoS and discuss the implementation of a new league for the robotics competitions: Junior Soccer Simulation league (JSS). Because soccer is the most popular sport in the world, we believe JSS will be a strong motivator for students to get involved with robotics.
Felipe N. Martins, Ivan S. Gomes, Carmen R. F. Santos

Path Planning with Collision Avoidance for Free-Floating Manipulators: A RRT-Based Approach

The difficulty of creating a path planner with collision avoidance for Space Manipulators (SMs) is well known due to the presence of dynamic singularities and because of its non-holonomic behaviour. Furthermore, the main contributions in the field of motion planning of SMs are often concentrated in the point-to-point strategy, with special interest in the complex dynamics of such systems. In fact, planners for space manipulators generally count on a previously computed path in order to modify it to avoid collisions. Nonetheless, the computing of the previous path still lacks robust formulations, specially in the case of free-floating manipulators. Our goal consists in creating a path planner with collision avoidance for a free-floating planar manipulator. The dynamic model is based on the Dynamically Equivalent Manipulator and the concept of Rapidly-Exploring Random Trees serves as a framework for the developed algorithm. A combination of a method that reduces the metric sensitivity with a bidirectional approach is proposed in order to achieve a solution convergence. Details of the collision checking algorithm are provided. The system is validated by simulating the path planning task for a three-link planar free-floating manipulator, while considering the presence of an obstacle. The results are then discussed and promising directions for future works are presented.
João R. S. Benevides, Valdir Grassi

A Topological Descriptor of Forward Looking Sonar Images for Navigation and Mapping

The automation of the monitoring, inspection and underwater maintenance tasks by underwater robots require a mapping and localization system. One challenge of these systems is how to recognize previously visited place in sensory information. This paper proposes a extended version of a method to detect loop closure dealing with acoustic images acquired by a forward looking sonar (FLS). The method builds a graph of Gaussian probability density function. This structure represents both shape and topological relation. We improve the image segmentation step adding a local parameters adjustment regard to intensity peak analyze of acoustic beams and changed the graph matching metric. We evaluate the method in a real dataset acquired by a underwater vehicle performing navigation in a harbor area.
Matheus Machado, Guilherme Zaffari, Pedro Ballester, Paulo Drews-Jr, Silvia Botelho

Non-stationary VFD Evaluation Kit: Dataset and Metrics to Fuel Video-Based Fire Detection Development

Datasets play a major role in the advance of computer vision techniques nowadays. Open, complete and challenging ground truth data, combined with standardized metrics are essential to push the development and allow the proper evaluation of computer vision algorithms. Even though a significant amount of work on VFD (video-based fire detection) systems has been developed, compare different algorithms is a laborious task due to the lack of common evaluation schemes and evaluation datasets. We address both of these issues by presenting a dataset of fire videos along with frame by frame annotations to be used for non-stationary fire detection algorithms training and validation. By the time, this is the largest dataset released on this subject matter. Standard video file formats and open markup languages where used to allow compatibility and convenient integration with the most popular computer vision libraries. The dataset includes hand-held, robot attached and drone attached footages and aims to boost the development of fully autonomous firefighter robots. The presented ground truth and metrics adapt to the majority of the state-of-the-art techniques and provides a reliable and unbiased solution to compare them. The dataset, example source-code and documentation are publicly available under the Creative Commons 3.0 license on GitHub.
Cristiano Rafael Steffens, Ricardo Nagel Rodrigues, Silvia Silva da Costa Botelho

GPU-Services: GPU Based Real-Time Processing of 3D Point Clouds Applied to Robotic Systems and Intelligent Vehicles

The GPU-Services project fits into the context of research and development of methods for data processing of three-dimensional sensors data applied to mobile robotics and intelligent vehicles. The implemented methods are called services on this project, which provide 3D point clouds pre-processing algorithms, such as, data alignment, segmentation of safe/unsafe navigable zones (e.g. separating ground from obstacles and borders/curbs) and elements of interest detection. Due to the large amount of data provided by the sensors to be processed in a very short time, these services use the GPU (NVidia CUDA) to perform partial or complete parallel processing of these data. The project aims to provide data processing services to an autonomous car, forcing the services to approach real-time processing, which is defined as completing all data processing routines before the arrival of the sensor’s next frame. This work was implemented considering 3D data acquired from a LIDAR, more specifically from a Velodyne HDL-32. The sensor data is structured in the form of a cloud of three-dimensional points, allowing for great parallel processing. However, the major challenge is the high rate of data received from this sensor (around 700,000 points/sec or 70.000 points/frame at 10 Hz), which gives the motivation of this project: to use the full potential of sensor and to efficiently use the parallelism of GPU programming. The GPU services are divided into four steps: The first step is an intelligent extraction, reorganization and spacial correction of the data provided by the Velodyne multi-layer laser sensor; The second stage is the segmentation of planar data; The third stage is object segmentation; The fourth stage is to develop a methodology that unite the results from the previous steps in order to better detect the curbs. The services were implemented and the performance was evaluated using traditional sequential data processing (CPU data processing) and parallel data processing (GPU CUDA implementations). Besides that, different NVidia GPUs were also tested, allowing us to process the acquired data much faster than using the CPUs, and in some cases faster than it was provided by the Velodyne sensor.
Leonardo Christino, Fernando Osório

Collaborative Object Transportation Using Heterogeneous Robots

The use of multi-robot systems can be seen in many different contexts in recent years. One of them is the object transportation problem, which has many applications, such as simple moving objects as well as in more complex scenarios, like tasks typically involved in building sites and structures assembling. Despite the fact that much effort has been focused on what may apparently be a relatively simple task, several facets of the problem still remain open and need to be tackled. In this work, we propose a complete methodology which encompasses all related stages of the problem (i.e. path planning, task allocation and control). Several experiments with simulated robots and with real ground robots were conducted in order to provide a thorough evaluation and validation of the methodology.
Ramon S. Melo, Douglas G. Macharet, Mario Fernando M. Campos

ND-NCD: Environmental Characteristics Recognition and Novelty Detection for Mobile Robots Control and Navigation

Mobile robot applications usually perform a path planning and its execution considers a previous known map. On the other hand, some application must explore the environment, defining a path from a source to a destination point, without knowing the environment map. The environment exploration, path planning towards a goal and navigation control tasks should be done at the same time. This study proposes a new method for mobile robot control and navigation based on the environmental characteristics recognition and novelty detection, named ND-NCD (Novelty Detection with Normalized Compression Distance). This method can be used as a key component in environment exploration and topological mapping tasks. In a previous work, a Genetic Algorithm (GA) for exploratory path planning was implemented to create a topological map (graph) from the source to the destination point, generating a set of actions which the robot must perform to achieve the goal. Each action was associated to a different reactive behavior specifically designed for characteristic places of the environment, such as corridors, curves or intersections. The proposed method, ND-NCD is used to recognize such different environmental characteristics, allowing to activate/associate the adequate actions whenever the method recognizes a context change (new context). This allows us to integrate the GA based environment exploration method together with the robot control reactive behaviors, which can be properly selected and switched according to the environmental characteristics detected/discovered by the ND-NCD. The ND-NCD uses the robot perception (e.g. laser sensor) to detect novelty and to recognize already known characteristics, thus allowing an incremental representation of the environment structures. The experiments were performed in the Player/Stage simulator and in a real indoor environment. ND-NCD performance is compared with a Neural Network trained to recognize context changes in the same environment. The results indicate that ND-NCD is a promising approach to be used in exploration and navigation control for mobile robots with the advantage of detecting a context change just knowing an initial state (corridor) from the environment. The proposed method does not need to be trained previously in order to know all the states (supervised training), being able to incrementally discover the different environment configurations.
Antônio Soares, Valéria Santos, Cláudio Toledo, Fernando Osório, Alexandre Delbem

Implementing and Simulating an ALLIANCE-Based Multi-robot Task Allocation Architecture Using ROS

In this chapter, we discuss the implementation and simulation results of a ALLIANCE-based architecture on Robot Operating System (ROS). In this approach, the system parameters were set empirically and we do not discuss system performance metrics. The focus is implementing the task allocation algorithm. After briefly review MRTA problem, we compare known architectures in some key aspects. Although only simulations validate the ALLIANCE-based approach, system flexibility and adaptivity is notable despite its runs variations.
Wallace Pereira Neves dos Reis, Guilherme Sousa Bastos

Humanoid Robot Gait on Sloping Floors Using Reinforcement Learning

Climbing ramps is an important ability for humanoid robots: ramps exist everywhere in the world, such as in accessibility ramps and building entrances. This works proposes the use of Reinforcement Learning to learn the action policy that will make a robot walk in an upright position, in a lightly sloped terrain. The proposed architecture of our system is a two-layer combination of the traditional gait generation control loop with a reinforcement learning component. This allows the use of an accelerometer to generate a correction for the gait, when the slope of the floor where the robot is walking changes. Experiments performed on a real robot showed that the proposed architecture is a good solution for the stability problem.
Isaac J. Silva, Danilo H. Perico, Thiago P. D. Homem, Claudio O. Vilão, Flavio Tonidandel, Reinaldo A. C. Bianchi

Structure-Control Optimal Design of 6-DOF Fully Parallel Robot

This contribution aims at introducing an optimal design methodology for the Stewart Platform robot that considers structure and control design variables simultaneously. This methodology intends to maximize the positioning accuracy in order to optimize the overall performance of the robot for a specific task. The structure design variables of the mechanism combined with the gains of the controller are the structure-control design variables, this global set is considered simultaneously in the optimal design methodology. A position control scheme, based on a PD controller, and the complete dynamics of the robot are considered to compute the overall tracking position as function of the structure-control design variables. A sensitivity analysis is performed to evaluate the effect of the structure-control design variables on the tracking position accuracy of the robot. The associated optimization problem is solved by using metaheuristic optimization methods. Simulation results demonstrate that the proposed design procedure is effective to increase the positioning accuracy, as well as to improve the closed loop dynamics performance of the robot.
Fabian Andres Lara-Molina, Didier Dumur, Edson Hideki Koroishi

A Genetic Algorithm Approach to the Automated System for Solving the Container Loading Problem

On the one hand the container loading problem has been widely studied in an effort to reduce logistical costs. On the other hand, automated planning research has as an objective assisting industrial processes by processing a system model and providing a list of actions that will enable the system to get from a given initial state to an objective. This works proposes an approach that combines CLP solving and automated planners to create a system that can execute the entire loading process. The CLP is solved by an improved genetic algorithm and its resulting packing pattern is converted to a format accepted by existing automated planners, whose output is a set of actions which can be executed to carry out the loading of the container.
Rodrigo Nogueira Cardoso, Marco Vinícius Muniz Ferreira, Alexandre Rodrigues de Sousa, José Jean-Paul Zanlucchi Souza Tavares

Trajectory Planning for UGV Using Clothoids

Path planning and autonomous navigation are the important challenges in mobile robotics. These are difficult tasks because the robot has to accurately and safely perform autonomous maneuverings. This work presents a methodology to plan the trajectory of a robot in dynamic and complex environments. Also, the changing lanes of one simulated car, which it traverse autonomously. A planner based in the AD* algorithm is used to plan a less costly trajectory to the destination for the task of automatic parking. For the changing lanes, we use the clothoid creation method, which is useful for avoid a vehicle in front of it. The methodology enables the robot to reach the goal, which is applied to determine the speed and steering of the robot. The results show that the methodology can create smooth clothoid trajectories to the vehicle follow.
Lucas P. N. Matias, Tiago C. Santos, Denis F. Wolf, Jefferson R. Souza


Weitere Informationen