ROBOT2013: First Iberian Robotics Conference

This paper presents an approach to perform data association in a monocular visual SLAM context. The proposed approach is designed to avoid the detection of false associations by means of RANSAC, and is well suited to help in localizing a robot in underwater environments. Experimental results embed the data association in a

trajectory-based SLAM

in order to evaluate its benefits when localizing an underwater robot. Qualitative and quantitative results are shown evaluating the effects of dead reckoning noise and the frequency of the SLAM updates.

This paper represents research in progress in autonomous manipulation for underwater intervention missions within the context of the GRASPER project. This project focuses on developing manipulation skills for an Autonomous Underwater Vehicle (AUV). Current research in underwater robotics intends to increase autonomy for all kinds of robotic intervention operations that require physical interaction. Very few underwater systems have the capacity to carry out intervention without any kind of umbilical cables for tele-operating the actions. This article aims to investigate new approaches to follow with the aforementioned challenges, with the inclusion of learning and probabilistic techniques to increase the autonomy levels of an underwater manipulation system. With this goal, a collaboration research action has been established between the IRS-Lab at UJI (Spain), as experts in the underwater robotic manipulation domain, and the Institute of Systems and Robotics from University of Coimbra (Portugal), experts in learning by interaction within a robotic manipulation context.

This paper presents the Squirtle, an Autonomous Surface Vehicle (ASV) for autonomous navigation and capable of manoeuvring in tight river environments. Its main features include precise localization with RTK GPS and inertial sensors, high manoeuvrability and autonomy, depth sensing and obstacle detection. Its purpose is the exploration and mapping of the bottom of rivers and of their shores. The article includes a brief description of the systems developed to fulfil this task. Software is based on the Robot Operating System (ROS). Results from preliminary field trials are presented and discussed.

This paper addresses the problems of Terrain Based Navigation (TBN) and Occupancy Grid Mapping for an Autonomous Underwater Vehicle (AUV). The two problems are solved using the same tools to make feasible in future works to implement a Simultaneous Localization and Mapping (SLAM). Realtime Occupancy Grid Mapping on the real vehicle Girona500 AUV is achieved by means of the Octomap library. The resulting map is later used for TBN with the parallelized execution of a Particle Filter making also use of the Octomap library to compare multibeam sonar ranges against the known map. The Occupancy Grid Mapping and the Particle filter are implemented as individual nodes in the vehicle’s software architecture in ROS. Tests were carried out in a dataset of a natural environment near the coast. Several parameters involving the Particle Filter (number of particles, number of beams, uncertainty of measurements) are studied. Finally, the results are compared with the dead reckoning obtained by the AUV and the USBL positions obtained from a surface boat.

There is considerable worldwide interest in the use of groups of autonomous marine vehicles to carry our challenging mission scenarios, of which marine habitat mapping of complex, non-structured environments is a representative example. Relative positioning and formation control becomes mandatory in many of the missions envisioned, which require the concerted operation of multiple marine vehicles carrying distinct, yet complementary sensor suites. However, the constraints placed by the underwater medium make it hard to both communicate and localise the vehicles, even in relation to each other, let alone maintain them in a formation. As a contribution to overcoming some of these problems, this paper deals with the problem of keeping an autonomous marine vehicle in a moving triangular formation with respect to two leader vehicles. Simple feedback laws are derived to drive a controlled vehicle to its intended position in the formation using acoustic ranges obtained to the leading vehicles with no knowledge of the formation path. The paper discusses the implementation of this solution in the MEDUSA class of autonomous marine vehicles operated by IST and describes the results of trials with these vehicles exchanging information and ranges over an acoustic network.

Two-dimensional forward-looking sonars are becoming standard sensors in both remotely operated and autonomous underwater vehicles, increasing the possibility of mapping under low visibility conditions. Due to the inherent nature of sonar image formation, the ideal mapping strategy relies on maintaining the same orientation, so as to minimize intensity alterations due to viewpoint changes. However, this is not always possible and therefore it is necessary to deal with the registration of sonar images under rotational movements. Previous investigations have discouraged the use of feature-based techniques and have suggested the use of global methods that are robust to noise, low-resolution and inhomogeneous insonification and can deal with the decoupled estimation of roto-translations. In this paper we review several candidate methods and assess them by using real data gathered under different conditions. By identifying the best approach for rotation estimation we aim to extend the applicability of sonar mosaicing to more diverse scenarios. Results indicate that applying phase correlation directly to polar frames leads to the highest accuracy under most cases.

This paper deals with the automatic computation of the assembly sequence for building truss structures from their 3D geometrical analysis. This functionality is part of the autonomous planning architecture of a team of aerial robots equipped with on-board robotic arms. The mission of the team is the construction of a structure in places where the access is difficult by conventional means. The assembly sequence is computed by applying the well known “assembly-by-disassembly” technique to the Non-Directional Blocking Graphs (NDBG) obtained from the geometrical analysis of the structure. In this paper two novel local heuristics are presented to solve the assembly problem: the former is based on the number of free nodes in the graphs and the latter is related to the size of the resulting connected subgraphs when each disconnection is applied to a set of parts. Both techniques are designed to compute the assembly sequence that allows to parallelize the building process of the structure if enough robots are available. Simulation results as well as experimental results with an aerial robot are presented in the paper.

In the last decade, the use of small autonomous unmanned aerial vehicles (UAVs) has been generalized. Specially, multirotor vehicles become really popular and are being employed in several different applications and fields: military missions, agricultural processes, rescue and surveillance operations, audiovisual productions.. and even have arrived to the public at large. This entertainment approach has set some challenges still unresolved: among them, the safety of these systems.

The work presented in this paper is framed in this challenge, focusing on the multirotors systems behaviour when facing a lift problem. It has been tried to detect and manage emergency situations, being the goal to minimize as much as possible the potential damages. It started from a deep study of the main malfunctions or breakdowns affecting to the drone’s lift. This study flowed into the design and develop of a model-based algorithm able to detect this events in a fast and robust way. This failure monitor allows to effectively apply control techniques to compensate (if it is possible) the breakdown’s effects and to use passive safety methods (e.g. parachutes) to minimize the potential damages derived from a fall.

The system and techniques developed in this work were tested both in simulation and in real experiments, proving they suppose and increment in the operation’s safeness (i.e. both the own drone and the third party agents, safety methods).

This paper describes the trajectory control of a quadrotor using external position estimation obtained from visual tracking in a scenario with multiple quadrotors and a camera mounted in the base of two or more UAVs, with the images being transmitted through a radio link. Applications where visual tracking can be used include fault detection and recovery of internal sensors, formation flying and autonomous aerial refueling. The dynamic model of the quadrotor and its trajectory control scheme is described along with the model of perturbations considered for the external position estimation. Graphical and numerical results are presented in different conditions, commenting separately the effect of each identified perturbation over the trajectory control. This study is done in simulation as previous step before testing quadrotor trajectory control in real conditions due to the high risk of accidents and damages on the vehicle.

The aim of this paper is to show a methodology to obtain a model of a rotary wing UAV (Unmanned Aerial Vehicle) employing a frequency-domain System Identification (SYSID) methodology using CIFER

®

. The methodology is applied to the CB-5000 RUAV and discuss several identification issues, from the telemetry acquisition process, parametric model to be identified and identification technique, to finally validate and implement the model. The UAV’s real autopilot software is integrated with the CIFER

®

model showing a good behaviour without any change on the tuning of the real autopilot gains. In order to validate and compare the results, an alternative two rigid body kinematic model is presented. Finally, the models integrated with the autopilot are compared by using the experimental data of the real RUAV (Rotorcraft UAV) platform following the same flight plan.

This work addresses the problem of fault detection and diagnosis (FDD) for a quad-rotor mini aerial vehicle (MAV). Actuator faults are considered on this paper. The basic idea behind the proposed method is to estimate the faults signals using the extended state observers theory. To estimate the faults, a polynomial observer is presented by using the available measurements and know inputs of the system. In order to investigate the observability and diagnosability properties of the system, a differential algebra approach is proposed. Furthermore, an evaluation function depending on the system states is developed, in order to be used in a controller, which will compensate the failures. The effectiveness of the methodology is illustrated by means of numerical simulations and some experimental tests.

Tremor is the most common movement disorder and strongly increases in incidence and prevalence with ageing. Although not life threatening, upper limb tremors hamper independent life of 65% of those suffering from them, greatly impacting on their quality of life. Current treatments of tremor include drugs and surgery. However, tremor is not effectively managed in 25% of patients. Therefore, further research and new therapeutic options are required for an effective management of pathological tremor. This paper introduces some rehabilitation robots developed for tremor suppression based on biomechanical loading, their evaluation and the identification of their limitations. At the end, authors aim to provide a view of the potential of this novel approach for tremor management and the plans for commercialization.

Robotic rehabilitation has become very popular in recent years. Nevertheless it still faces various problems related to the system actuation devices. Some classic actuator features, like the noise, weight, size-force relationship, and the efficiency, make them little suitable for such applications. For this reason, piezoelectrically driven ultrasonic motors have become the alternative actuators to the conventional electromagnetic motors. This is due to their light weight, compact size and soundless performance characteristics. The aim of this article is to present a new actuator based on an ultrasonic motor (USM) and a magnetorheological clutch is integrated into a device used in robotic rehabilitation for the elbow joint. At the end, a control strategy algorithm for the USM and its first preliminary results will be presented. The work presented was carried at the Systems Engineering and Automation Department of Carlos III University of Madrid, under the HYPER CONSOLIDER-INGENIO 2010 Spanish project.

This article presents the skeletal-mathematical model and a measurement system for providing real-time information related to position, angular velocity, and orientation of different movements related to the upper limb of human shoulder using Inertial Measurement Units (IMU). The main goal of this system is to improve the rehabilitation tasks of unilateral brachial plexus injury, therefore a complete kinematic analysis of the shoulder skeletal system, considering only the essential variables to simulate the movements used in rehabilitation, is detailed. Finally, are presented some experimental tests based on joint physiology and biomechanics of human shoulder for children less than 10 years, according to desired restriction of medical staff belong to Hospital Infanta Sofía of Madrid.

This paper presents a multi-modal interface for interaction between people with physical disabilities and an assistive robot. This interaction is performed through a dialogue mechanism and augmented 3D vision glasses to provide visual assistance to an end user commanding an assistive robot to perform Daily Life Activities (DLAs). The augmented 3D vision glasses may provide augmented reality vision of menus and information dialogues over the view of the real world, or in a simulator environment for laboratory tests and user evaluation. The actual dialogue is implemented as a finite state machine, and includes possibilities of Automatic Speech Recognition (ASR), and a Text-to-Speech (TTS) converter. The final study focuses on studying the effectiveness of these visual and auditory aids for enabling the end user to command the assistive robot ASIBOT to perform a given task.

Lower-limb exoskeletons and powered orthoses in gait assistance applications for patients with locomotive disorders possess the potential to significantly affect society in the near future. This paper presents the primary features of a lower-limb exoskeleton to enable paralysed children to walk. Because these patients are unable to move their limbs, the device generates their basic motions in everyday life, e.g., standing up, sitting down, and stable ambulation. A walker provides stability in the lateral plane, while the active orthosis provides stability in the sagittal plane while walking. The walker has been devised with a two degree of freedom mechanism to allow the user to sit down and stand up in a stable and comfortable way without the movement of the walker itself. The gait of the orthosis parameters such as step height, body height or step length are modified online, based on an impedance control approach, providing a safe and smooth gait pattern. Two shoe insole pressure measurement systems provide ground reaction force and center of pressure to adapt these gait parameters online. An adjustable compliance actuator has been designed and incorporated to the knee joint of the active orthosis. This Actuator with Adjustable Rigidity and Embedded Sensor (ARES) fulfills the demanding characteristics required in an active orthosis’s joint, namely, intrinsic compliance to allow human-machine interaction, high power-to-weight ratio, high peak torque, small size and low weight. Exploiting the characteristics of ARES actuator a control scheme has been designed and implemented to achieved a reduction in the energy expenditure while keeping compliant to accommodate unexpected disturbances. The final ATLAS exoskeleton has been successfully tested in a healthy user, in a quadriplegic child, and in a patient with neuromuscular disease.

Heart motion compensation is a challenging problem within medical robotics and it is still considered an open research area due to the lack of robustness. As it can be formulated as an energy minimization problem, an optimization technique is needed. The selection of an adequate method has a significant impact over the global solution. For this reason, a new methodology is presented here for solving heart motion compensation in which the central topic is oriented to increase robustness with the goal of achieving a balance between efficiency and efficacy. Particularly, genetic algorithms are used as optimization technique since they can be adapted to any real application, complex and oriented to work in real-time problems.

Endowing current surgical robotic systems with haptic feedback to perform minimally invasive surgery (MIS), such as laparoscopy, is still a challenge. Haptic is a feature lost in surgical teleoperated systems limiting surgeons capabilities and ability. The availability of haptics would provide important advantages to the surgeon: Improved tissue manipulation, reducing the breaking of sutures and increase the feeling of telepresence, among others. To design and develop a haptic system, the measurement of forces can be implemented based on two approaches: Direct and indirect force sensing. MIS performed with surgical robots, imposes many technical constraints to measure forces, such as: Miniaturization, need of sterilization or materials compatibility, making it necessary to rely on indirect force sensing. Based on mathematical models of the components involved in an intervention and indirect force sensing techniques, a global perspective on how to address the problem of measurement of tool-tissue interaction forces is presented.

Surgical robotics has usually provided the handle of surgical tools by using teleoperated systems or the automation of certain surgical tasks. However, the final goal of this field has always consisted of allowing the surgeon to perform an intervention without additional human assistance (i.e. solosurgery). In this way, this paper is focused on the design and implementation of a semi-autonomous surgical robot capable of assisting the surgeon during a suture procedure on a Hand Assisted Laparoscopic Surgery (HALS) scenario. The methodology proposed is based on a dispatcher system which manages the actuations of the robot depending on the recognized gestures of the surgeon’s tool. The performance of the whole architecture has been tested by means of in vitro trials.

This paper presents an implementation of the Multiple Hypothesis Tracking (MHT) algorithm in the Advanced Driver Assistance Systems (ADAS) context.The proposed algorithmuses laser data received from two front mounted sensors on a mobile vehicle. The algorithm was tested with simulated and real world data and shown to obtain a very good performance. Nonholonomic motion models were used to model the movement of road agents instead of the more traditional constant velocity/acceleration models. The use of the nonholonomic motion models allows to obtain not only the linear velocity, but also the steering angle of vehicles, improving this way the future prediction and handling of occlusions. The MHT algorithm possesses some well-known critical disadvantages due to its complexity and computational growth, in this work we circumvent these limitations in order to achieve real time performance in real work conditions.

Programming by Demonstration (PbD) covers methods by which a robot learns new skills through human guidance and imitation. PbD has been a key topic in robotics during the last decade that includes the development of robust algorithms for motor control, motor learning, gesture recognition and the visual-motor integration. Nowadays, PbD deals more with learning methods than traditional approaches, and frequently it is referred to as Imitation Learning or Behavioral Cloning. This work will review and analyse existing works in order to create a taxonomy of the elements that constitute the most relevant approaches in this field to date. We intend to establish the categories and types of algorithms involved so far in PbD and describing their advantages and disadvantages and potential developments.

The rising number of applications involving service robots demands more specific and custom control strategies. Under those special working conditions on-road signaling and beaconing robots are framed. Control strategies that transcend each individual robot and focus on the whole set are required, instead of treating each one as a single entity. Thus, control methodology - which usually involves both planned control as behavioral based control - will depend not only on each single robot’s parameters, but also on the parameter analysis of other group robots, taking them into account within the decision process. This paper describes an approach to a collaborative control architecture that enhances the mapping and positioning tasks of the robot, as well as the functional organization of the whole robot group. That enhancement is achieved by means of mutual recognition between system’s robots, the use of specific sensors and the sharing of robot’s position data together with its movement parameters and the information about its surroundings.

The present paper presents the ongoing research undertaken as a research experiment, called HERMES, inside Echord european integrated project framework. The goal of the HERMES experiment is to study, analyze and finally implement the packaging of shoes on a robot system that mimics the required degree of flexibility and dexterity provided by the human workers. The use of a bimanual system with anthropomorphic hands has been chosen as it could also be applied to solve other processes of similar or even higher complexity in the future and thus provide a holistic approach towards automization in footwear industry. In this paper, the first experimental results of the HERMES experiment are presented.

In this article we propose a method for cooperatively building a scene map between a human and a robot by using a spatial relational model employed by the robot to interpret human descriptions of the scene. The description will consist in a set of spatial relations between the objects in the scene. The scene map will contain the position of these objects. For this end we propose a model based on the generation of scalar fields of applicability for each of the available relations.

The method can be summarized as follows. In first place a person will come into the room and describe the scene to the robot, including in the description semantic information about the objects which the robot can’t get from its sensors. From the description the robot will form the “scene mental map”. In second place the robot will sense the scene with a 2D range laser building the “scene sensed map”. The objects positions in the mental map will be used to guide the sensing process. In a third step the robot will fuse the two maps, linking the semantic information about the described objects to the corresponding sensed ones. The resulting map is called the “scene enriched map”.

An important problem in robotics is the empirical evaluation of classification algorithms that allow a robotic system to make accurate categorical predictions about its environment. Current algorithms are often assessed using sample statistics that can be difficult to interpret correctly and do not always provide a principled way of comparing competing algorithms. In this paper, we present a probabilistic alternative based on a Bayesian framework for inferring on balanced accuracies. Using the proposed probabilistic evaluation, it is possible to assess the balanced accuracy’s posterior distribution of binary and multiclass classifiers. In addition, competing classifiers can be compared based on their respective posterior distributions. We illustrate the practical utility of our scheme and its properties by reanalyzing the performance of a recently published algorithm in the domain of visual action detection and on synthetic data. To facilitate its use, we provide an open-source MATLAB implementation.

In this work is presented a general architecture for a multi physical agent network system based on the coordination and the behaviour management. The system is organised in a hierarchical structure where are distinguished the individual agent actions and the collective ones linked to the whole agent network. Individual actions are also organised in a hybrid layered system that take advantages from reactive and deliberative control. Sensing system is involved as well in the behaviour architecture improving the information acquisition performance.

This paper presents a real-time vision framework that detects and tracks vehicles from stationary camera. It can be used to calculate statistical information such as average traffic speed and flow as well as in surveillance tasks. The framework consists of three main stages. Vehicles are first detected using Haar-like features. In the second phase, an adaptive appearance-based model is built to dynamically keep track of the detected vehicles. This model is also used in the third phase of data association to fuse the detection and tracking results. The use of detection results to update the tracker enhances the overall framework accuracy. The practical value of the proposed framework is demonstrated in real-life experiments where it is used to robustly compute vehicle counts within certain region of interest under variety of challenges.

The objective of the work presented in this paper is to investigate the optimal learning strategy for snake-like modular robots using a (1+1) Evolutionary Algorithm. We take into account three different but correlated tasks: efficient locomotion, reaching a given point and obstacle avoidance.

Starting from earlier results on locomotion, we have performed three different sets of experiments. In the first, the snake must learn to go to a goal point, and we investigate how employing different fitness functions affect the learning of this task. In the second experiment the snake must learn how to avoid obstacles. In this experiment we test two possible behaviors, called

mixed

and

switched

strategies. Finally, in the third set of experiments, we introduce the concept of

incremental learning

and compare it with the “all-at-once” learning schemes of the first two experiments.

The results of the simulations indicates that the modular robots are able to learn both tasks with the (1+1) Evolutionary Strategy adopted, and that the fitness function that explicitly rewards each of the tasks perform better that the fitness function that takes into account the locomotion task only implicitly, rewarding only the reaching of the target point. We also demonstrate that the obstacles avoidance configuration with only one behaviour (mixed) is better than the configuration with two behaviours (switched) and that incremental learning provide a faster evolution towards good controllers.

Applications with large teams of robots are becoming more and more useful. If the scenario is very crowded or very dynamic, conflict resolution when using a shared workspace is a challenging problem. In this paper, an scalable, decentralized and reactive approach for collision avoidance is presented. The robots can navigate in a 2D environment avoiding each other and without high computational requirements. In addition to the conflict resolution algorithm, a multi-robot simulator is presented. The system is flexible and can be used to simulate different algorithms with realistic robots. Finally, an extension of the simulator is proposed in order to operate real robots in a multi-robot testbed. Results of the collision avoidance approach are shown with both real and simulated robots.

This research presents how augmented reality (AR) can be used for assistance tasks in indoor environments. The approach of AR supported human-robot interaction arises two classical questions in robot navigation: What does the robot need to do to show the AR? (task planning) and, which are the best movements to reach the destination allowing the user to get the AR information? (path planning). In addition, a constraint related to assistance environments has to be considered: How can the robot perform its tasks in order to improve the interaction?. These problems along with how AR can enhance human-robot interaction, and therefore improve the human experience with the robot, are addressed along the paper. The first experiments developed with this system are performed using the gazebo simulator. The solutions proposed are analized using a low cost platform for elderly assistance, which already uses AR for some assistance tasks, developed in Universidad de Len.

People detection and tracking is an essential skill to obtain social and interactive robots. Computer vision has been widely used to solve this task but images are affected by noise and illumination changes. Laser range finder is robust against illumination changes so that it can bring useful information to carry out the detection and tracking. In fact, multisensor approaches are showing the best results. In this work, we present a new method to detect and track people using a laser range finder. Patterns of leg are learnt from 2d laser data using supervised learning. Unlike others leg detection approaches, people can be still or moving at the surroundings of the robot. The method of leg detection is used as observation model in a particle filter to track the motion of a person. Experiments in a real indoor environment have been carried out to validate the proposal.

Precision farming goal is to maximize the productivity of the crops while minimizing the use of resources. Given that the agricultural technique has evolved along the millenniums, the only real option to keep increasing the agricultural productivity requires the use of technology: currently, many different robots are being used to harvest, plow and identify weeds. This work aims to reduce the environmental invasion the use of these implies, maximizing at the same time the knowledge of the crop status. We propose a new type of environmental-friendly ground mobile robot employed to monitor temperature and humidity in agricultural fields Used as a distributed sensor web. It takes measures of these environmental variables in any specific place required. It also allows to determine zones of the crop with bad irrigation or under germination risks and take corrective actions on time, providing better growing conditions: adequate germination time, faster growing and better efficiency. We present its architecture and a new external module used to measure the temperature and humidity. Beside, we also present two experiments carried out to validate the spherical robot approach: the results obtained first at a cornfield and the second in small orchard in CAR-UPM-CSIC facilities are exposed and analysed.

The RHEA project is based on the cooperation among aerial and ground vehicles aiming at performing several precision agriculture tasks. This work is focused on the description of the aerial mission in the mentioned project. The work includes a description of the units, the goals and mission types, the requirements involved in the mission planning as well as their supervision and monitoring. The missions are performed by using a fleet of last-generation hex-rotors that rely on high payload and extraordinary stability. These features allow taking steady pictures with high quality cameras in large extension fields. As result, high-resolution images of the field to cover are obtained in order to provide with weeds pots positions to ground units for their removal.

Specific weed management consists on adjusting herbicide treatments depending on the zone infested and the type of weed. In this context, the discrimination between grasses (monocots) and broad-leaved weeds (dicots) is an important objective mainly because the two weed groups can be appropriately controlled by different specific herbicides. This work proposes a method of discrimination between these types of weeds based on a combined strategy, the Sugeno Fuzzy Integral, where the final decision is taken by combining seven attributes, the Hu moments. The main challenge in terms of image analysis is to achieve an appropriate discrimination between both groups in outdoor field images under varying conditions of lighting as well as of soil background texture.

This paper presents the main characteristics of a robot whose aim is to perform field inspection using autonomous navigation. The solution developed for crop row tracking is shown, which is a fundamental behaviour for crop inspection. For this purpose, an image processing method is implemented to determine the vehicle’s relative position to the crop row in real time. This position is supplied to two fuzzy controllers, one for angular speed and the other for linear speed. To integrate crop row tracking and other skills that the robot needs, we propose generating the different behaviours of the robot using a network of nodes with different functions: perceptive nodes, cognitive nodes and actuator nodes. The actions of the robot emerge from this set of behaviours, depending on the goals and needs that must be met at each given moment in time.

This paper presents a multisensory system for the detection and localisation of fruits that are candidates to be harvested by a robotic manipulator. The devices that have been selected as primary sensors for this purpose are a high resolution colour camera, a multispectral imaging system that consists of a motorised filter wheel, and a Time-Of-Flight 3D camera. A controlled lighting system completes the set-up. The progressive RGB camera and the multispectral imaging system acquire the basic data inputs for the detection of areas of interest that belong to the fruits, whereas the Time-Of-Flight 3D camera provides fast acquisition of accurate distances enabling the localisation of the targets in the coordinate space. Several experimental tests have been carried out in laboratory conditions in order to evaluate the capabilities of the proposed multisensory system.

Lately, many computer-based sensors and actuators have been incorporated in agricultural equipment with the main goal of configuring agricultural robots capable of achieving different tasks autonomously. However, the incorporation of different electronic systems in a robot impairs its reliability and increases its cost. Thus, hardware minimization and ease of software integration is mandatory to obtain feasible robotic systems. This paper strives to find a hardware architecture for both individual robots and robots working in fleets to improve reliability, decrease development costs and allow the integration of software from different developers.

REEM is a humanoid service robot designed to provide useful applications in public spaces and to assist people in domestic environments. REEM is the result of several years of research in real scenarios. Its main functionalities are autonomous navigation, person detection and recognition, speech synthesis and recognition.

This paper proposes a locomotion control system for biped robots by using a network of Central Pattern Generators (CPGs) implemented with Matsuoka’s oscillators. The proposed control system is able to control the system behaviour with a few parameters by using simple rhythmical signals. A network topology is proposed in order to control the generation of trajectories for a biped robot in the joint-space both in the sagittal and coronal planes. The feedback signals are directly fed into the network for controlling the robot’s posture and resetting the phase of the locomotion pattern in order to prevent the robot from falling down whenever a risk situation arises. A Genetic Algorithm is used to find optimal parameters for the system in open-loop. The system behaviour in closed-loop has been studied and analysed through extensive simulations. Finally, a real NAO humanoid robot has been used in order to validate the proposed control scheme.

Robots detect and keep track of relevant objects in their environment to accomplish some tasks. Many of them are equipped with mobile cameras as the main sensors, process the images and maintain an internal representation of the detected objects. We propose a novel active visual memory that moves the camera to detect objects in robot’s surroundings and tracks their positions. This visual memory is based on a combination of multi-modal filters that efficiently integrates partial information. The visual attention subsystem is distributed among the software components in charge of detecting relevant objects. We demonstrate the efficiency and robustness of this perception system in a real humanoid robot participating in the RoboCup SPL competition.

Humanoid robots are expected to work and collaborate with humans performing in changing environments. Developing this kind of robots requires them to display intelligent behaviors. For behaviours to be considered as intelligent they must at least present the ability to learn skills, represent skill’s knowledge, and adapt and generate new skills. In this work a framework is proposed for the generation and adaptation of learned models of robot skills for complying with task constraints. The proposed framework is meant to allow: for an operator to teach and demonstrate to the robot the motion of a task skill it must reproduce; to build a knowledge base of the learned skills knowledge allowing for its storage, classification and retrieval; to adapt and generate learned models of a skill, to new context, for compliance with the current task constraints. A learning from demonstration approach is employ to learn robot skill by means of probabilistic methods, encoding the motion dynamics in a

Gaussian Mixture Model

. We propose that this models of the skill can be operate and combine to represent and adapt the robot skills.

Variable Stiffness Actuators (VSAs) emerged as an alternative to conventional actuators in a variety of applications, such as walking robots and service robotics. New requirements, which were obviated in the design of rigid actuators, must be accounted for during the mechanical design of such new devices. Among them, we find the possibility of tunning the natural frequency of the mechanical system or damage reduction in case of impacts A multitude of solutions have been already proposed in the literature, each characterized by the kind of mechanism in charge of implementing the equilibrium position and the stiffness of the joint. With this work, we introduce a new actuator design based on tendon transmission, where a main motor controls the equilibrium position of the link while a secondary motor is in charge of modifying the joint stiffness. Unlike existing actuators, our proposal achieves a wide range of stiffness values, from close to zero up to completely rigid. Another distinguishing feature of the new design is the existence of a torque threshold, such that variable stiffness only becomes effective once the load is above a certain predefined threshold.

Exploring the full potential of humanoid robots requires their ability to learn, generalize and reproduce complex tasks that will be faced in dynamic environments. In recent years, significant attention has been devoted to recovering kinematic information from the human motion using a motion capture system. This paper demonstrates and evaluates the use of a Kinect-based capture system that estimates the 3D human poses and converts them into gestures imitation in a robot. The main objectives are twofold: (1) to improve the initially estimated poses through a correction method based on constraint optimization, and (2) to present a method for computing the joint angles for the upper limbs corresponding to motion data from a human demonstrator. The feasibility of the approach is demonstrated by experimental results showing the upper-limb imitation of human actions by a robot model.

In this paper we present a tool designed to provide entertainment by using social robots, specifically NAO humanoid robots. The tool takes as input a declarative representation of a theater piece that includes both text and movements to be said (performed) by robots. It can involve a one-robot monologue or a dialogue between two robots. Given the declarative representation of the input, it allows for anyone to create new pieces and easily test how they are represented by real or by virtual NAO robots.

This paper presents the novel humanoid postural control (PC) architecture for the humanoid robot TEO. It is outlined the high level and human inspired system for improving task performance. The study of the human PC system has inspired all processes involved in the control system. The information coming from sensors is interpreted applying neurophysics concepts and, then, the resulting perceptual parameters are applied for task performance improvement. The new PC system is an anticipative module complementing an existing reactive subsystem. This design tries to replicate the operation of the human case. In this way, the reactions can be more complex and higher perturbations levels can be overcome.

In this paper we analyze the issues surrounding the European Proposal for General Data Protection Regulation and the adaptation of law to the technical capabilities that have an impact on privacy. We start by analyzing the new capabilities of robotic machines and the involvement of new stakeholders as engineers, designers and manufacturers. Then we present the challenges of M2M relationships and the risks of being detected through devices. Next, we discuss the need to incorporate some concepts such as ‘pseudonym”, anonymous processing and legal profiling. We also argue that law needs technical standards and we foresee that engineers, designers and manufacturers will not have a neutral role. Future regulation should ensure some guidelines for privacy safe robots, and guarantee privacy and freedom in a technological world.

In this article we study the safety issue of deploying robots from the legal point of view. We start describing the needs for deploying robots and then the legal issues related to safe robots. Then we study the liability from the contractual and extra contractual point of views, and we explain how is the present legislation about the responsibility of the problems that can happen with robots and their deployment. Also, we discuss the company and public administration liability (quality problem) and how it is approached by the European legislation. Moreover, we discuss how tele-operation of robots by humans can affect the safety issues.

The development achieved by scientific and technical knowledge applied to robotics has long posed important ethical questions. In this paper are addressed some of these old and new questions that creates the human-robot relationship and its social dimension. This reflection is applied to the military use of drones, at a time in which the ethical and legal questions accompany some of these uses. As well as the plurality of evaluations determines the design and use of robots, the axiology is used to reflect the ethical problems of robotics.

In visual simultaneous location and mapping (SLAM) with a single camera, the use of 3D points as a basic feature has been shown sufficient to reliably estimate the camera position and orientation. Nevertheless, the resultant maps are not clear enough for certain applications, even for a large amount of point features. We propose a novel SLAM technique that uses lines as basic features, and the unscented Kalman filter (UKF) as a tracking algorithm. This paper discusses the mathematical foundations as well as the practical implementation of this technique, along with the results of preliminary experiments.

Event-based visual servoing is a recently presented approach that performs the positioning of a robot using visual information only when it is required. From the basis of the classical image-based visual servoing control law, the scheme proposed in this paper can reduce the processing time at each loop iteration in some specific conditions. The proposed control method enters in action when an event deactivates the classical image-based controller (i.e. when there is no image available to perform the tracking of the visual features). A virtual camera is then moved through a straight line path towards the desired position. The virtual path used to guide the robot improves the behavior of the previous event-based visual servoing proposal.

This paper proposes an appearance-based method to detect loop closure in visual SLAM (Simultaneous Localization and Mapping). To solve this problem, we make use of omnidirectional images and the internal odometry captured by a robot in a real indoor environment. We build an appearance-based model and, subsequently, two maps of the environment are constructed, one metric and other topological with relationships between them. These relationships are updated in each step of our hybrid approach. The topological map is a graph built from the appearance information in the scenes. A new node is added when the new visual information is different enough from the previous information. At the same time, we check a possible topological loop closure with previous nodes. On the other hand we estimate the metric position of the new pose using a Monte-Carlo approach with the aim of building a metric map. The experimental results demonstrate the reasonable performance of our method.

The problems of convergence to a desired configuration for a set of robots and leader-following in formation are considered in a framework where the robots have nonholonomic constraints, move in a plane and are observed by a calibrated flying camera, which provides the only sensory information used for the control. We propose a homography-based visual control method only requiring a priori single image of the desired configuration to perform the task. The proposed method consists of an image-based control scheme using the homography induced by the multi-robot system. Therefore, an interesting property is that the whole information regarding the multi-robot system is encapsulated in one single homography. The results show that the system is able to track the leader with the robots in formation despite the leader and camera motion are unknown.

Tracking by detection methods are becoming increasingly popular in recent years. They use samples classified in previous frames to detect object in a new frame. These methods have shown successful results. However, due to the self updating nature of this approach, tracking by detection methods usually suffer from object drift. Inaccurately detected samples are added to the training set which degrades the performance. Another problem is that the object may change in shape and size which increases the potential for inaccurate detection and subsequently the chance of losing the object. We propose a robust object tracking algorithm that adapts to changes in the size of the object. The algorithm is divided into two steps. The first step uses random projections compressed using a sparse matrix to describe positive and negative samples generated around the object. These samples are used to train a classifier to detect the position of the object in the next frame. The second step searches for the best size to fit the object around the position chosen in the first step. Experimental results show that our method provides robust tracking and help alleviate the drift problems.

This paper shows a performance comparison of two sensors capable of obtaining depth information using two different methods, i.e. stereo information and infrared based depth measurement. The sensors are a Bumblebee XB3 and a Microsoft Kinect, and they provide in-depth information with some advantages and disadvantages that will be presented and evaluated in this paper. The analysis compares the devices single characteristics and tests their performance.