Invited Speakers (alphabetically)
Dr. Ajoudani, Arash, Human-Robot Interfaces and Physical Interaction, Istituto Italiano di Tecnologia, Italy.
Title: "Interaction Control of Collaborative Robots for Adaptation to Task and Human Factors"
Abstract: The great potential and benefits of collaborative robots are becoming increasingly evident among both the scientific and industrial communities. Robots can reduce stress by providing assistance; increase human capabilities while preventing injuries in industrial scenarios, automatize therapies in rehabilitation contexts, and improve in general the quality of life. On the other hand, humans can provide experience, transfer knowledge and supervise robots’ functionalities, adding a certain level of adaptability to the process and contributing to an effective accomplishment of a broad range of tasks. Towards the improvement of robot adaptability to the variations of the task and human states, this talk will introduce novel control frameworks that are capable of processing the human states and task performance, and reacting accordingly.
Title: "Interaction Control of Collaborative Robots for Adaptation to Task and Human Factors"
Abstract: The great potential and benefits of collaborative robots are becoming increasingly evident among both the scientific and industrial communities. Robots can reduce stress by providing assistance; increase human capabilities while preventing injuries in industrial scenarios, automatize therapies in rehabilitation contexts, and improve in general the quality of life. On the other hand, humans can provide experience, transfer knowledge and supervise robots’ functionalities, adding a certain level of adaptability to the process and contributing to an effective accomplishment of a broad range of tasks. Towards the improvement of robot adaptability to the variations of the task and human states, this talk will introduce novel control frameworks that are capable of processing the human states and task performance, and reacting accordingly.
Prof. Billard, Aude, Learning algorithms and systems Laboratory, École polytechnique fédérale de Lausanne, Switzerland.
Title: "Variable impedance modeling through time-invariant dynamical system"
Abstract: Time-invariant dynamical systems have traditionally been used to provide kinematic control by generating desired reference trajectories. This talk presents different approaches to embed varying impedance profile by shaping the dynamical system and their application to perform contact tasks.
Title: "Variable impedance modeling through time-invariant dynamical system"
Abstract: Time-invariant dynamical systems have traditionally been used to provide kinematic control by generating desired reference trajectories. This talk presents different approaches to embed varying impedance profile by shaping the dynamical system and their application to perform contact tasks.
Dr. Calinon, Sylvain, Robot Learning & Interaction Group, Idiap Research Institute, Switzerland.
Title: "Challenges in extending learning from demonstration to variable impedance skills"
Abstract: Human-centric robotic applications often require the robots to learn new skills by interacting with the end-users. From a machine learning perspective, the challenge is to acquire skills from only few interactions, with strong generalization demands. It requires the development of intuitive active learning interfaces to acquire meaningful demonstrations, the development of models that can exploit the structure and geometry of the acquired data in an efficient way, and the development of adaptive control techniques that can exploit the learned task variations and coordination patterns. The developed models often need to serve several purposes (recognition, prediction, synthesis), and be compatible with different learning strategies (imitation, emulation, exploration). For the reproduction of skills, these models need to be enriched with force and impedance information to enable human-robot collaboration and to generate safe and natural movements. I will present an approach combining model predictive control and statistical learning for learning controllers exploiting the torque-controlled capability of robots. The proposed approach will be illustrated in various applications, with robots either close to us (robot for dressing assistance), part of us (prosthetic hand control from EMG and tactile sensing), or far from us (teleoperation with haptic feedback of bimanual robot in deep water).
Title: "Challenges in extending learning from demonstration to variable impedance skills"
Abstract: Human-centric robotic applications often require the robots to learn new skills by interacting with the end-users. From a machine learning perspective, the challenge is to acquire skills from only few interactions, with strong generalization demands. It requires the development of intuitive active learning interfaces to acquire meaningful demonstrations, the development of models that can exploit the structure and geometry of the acquired data in an efficient way, and the development of adaptive control techniques that can exploit the learned task variations and coordination patterns. The developed models often need to serve several purposes (recognition, prediction, synthesis), and be compatible with different learning strategies (imitation, emulation, exploration). For the reproduction of skills, these models need to be enriched with force and impedance information to enable human-robot collaboration and to generate safe and natural movements. I will present an approach combining model predictive control and statistical learning for learning controllers exploiting the torque-controlled capability of robots. The proposed approach will be illustrated in various applications, with robots either close to us (robot for dressing assistance), part of us (prosthetic hand control from EMG and tactile sensing), or far from us (teleoperation with haptic feedback of bimanual robot in deep water).
Dr. Nemec, Bojan, Department of Automation, Robotics and biocybernetics, Jožef Stefan Institute, Slovenia.
Title: "Impedance learning and control during bi-manual human robot cooperation"
Abstract: In this talk, we introduce an approach for intuitive and natural physical human-robot interaction in cooperative tasks, where we applied speed-scaled dynamic motion primitives for the underlying task representation. Through initial learning by demonstration, robot behavior naturally evolves into a cooperative task, where the human co-worker is allowed to modify both the spatial course of motion as well as the speed of execution at any stage. The robot adjusts its compliance in path operational space, defined with a Frenet-Serret frame. Furthermore, the required dynamic capabilities of the robot are obtained by decoupling the robot dynamics in operational space, which is attached to the desired trajectory. This allows a human co-worker in a cooperative task to be less precise in parts of the task that requires high precision, as the precision aspect is learned and provided by the robot. The user can also freely change the speed and/or the trajectory by simply applying force to the robot.
This approach was utilized also for learning by demonstration of bimanual tasks, where the operator is allowed to incrementally refine the desired policy by pushing the robot forward and backward along the motion trajectory and change the trajectory only where necessary. For assembly tasks, it is necessary to determine also the environmental constraints during the demonstration and use them to adjust the dynamics of the robot during the task execution. For this purpose, we utilize the principal component analyses framework. The last part of the talk will discuss autonomous learning of environment constraints applied to the tasks such as door and drawer opening.
Title: "Impedance learning and control during bi-manual human robot cooperation"
Abstract: In this talk, we introduce an approach for intuitive and natural physical human-robot interaction in cooperative tasks, where we applied speed-scaled dynamic motion primitives for the underlying task representation. Through initial learning by demonstration, robot behavior naturally evolves into a cooperative task, where the human co-worker is allowed to modify both the spatial course of motion as well as the speed of execution at any stage. The robot adjusts its compliance in path operational space, defined with a Frenet-Serret frame. Furthermore, the required dynamic capabilities of the robot are obtained by decoupling the robot dynamics in operational space, which is attached to the desired trajectory. This allows a human co-worker in a cooperative task to be less precise in parts of the task that requires high precision, as the precision aspect is learned and provided by the robot. The user can also freely change the speed and/or the trajectory by simply applying force to the robot.
This approach was utilized also for learning by demonstration of bimanual tasks, where the operator is allowed to incrementally refine the desired policy by pushing the robot forward and backward along the motion trajectory and change the trajectory only where necessary. For assembly tasks, it is necessary to determine also the environmental constraints during the demonstration and use them to adjust the dynamics of the robot during the task execution. For this purpose, we utilize the principal component analyses framework. The last part of the talk will discuss autonomous learning of environment constraints applied to the tasks such as door and drawer opening.
Dr. Howard, Matthew J. W., Department of Informatics, King's College London, UK.
Title: "Transferring Impedance Behaviours Combining Human and Robot Best Practice"
Abstract: The optimality of human impedance behaviour is tightly coupled to the specific properties of human embodiment. While this results in impressive in versatility for task oriented behaviour, it also imposes a puzzle on those wishing to imitate such behaviour in robotic systems. Robots can be designed differently, and therefore do not have the same restrictions on their control strategies, but the challenge then is to devise an appropriate variable impedance profile for a given task. This talk will discuss recent work at King’s College London, where we have been looking at ways to unpick this puzzle, through a combination of algorithmic approaches for blending human and robot ‘best practice’ and new ways to gain ubiquitous measurements of musculoskeletal behaviour.
Title: "Transferring Impedance Behaviours Combining Human and Robot Best Practice"
Abstract: The optimality of human impedance behaviour is tightly coupled to the specific properties of human embodiment. While this results in impressive in versatility for task oriented behaviour, it also imposes a puzzle on those wishing to imitate such behaviour in robotic systems. Robots can be designed differently, and therefore do not have the same restrictions on their control strategies, but the challenge then is to devise an appropriate variable impedance profile for a given task. This talk will discuss recent work at King’s College London, where we have been looking at ways to unpick this puzzle, through a combination of algorithmic approaches for blending human and robot ‘best practice’ and new ways to gain ubiquitous measurements of musculoskeletal behaviour.
Prof. Kheddar, Abderrahmane, CNRS & CNRS-AIST, France.
Title: "Contact detection for humanoids in close proximity interaction"
Abstract: Humanoid robots can serve as human partners in various close-contact situations. As real-use application perspectives appeared recently (domotics, large-scale manufacturing…), humanoids have great potential to be exploited as sophisticated assistive or cobotic systems. Their shape imparts them interesting properties in terms of integration, interaction with humans, empathy, and embodiment. As a key element in human-humanoid physical interaction is the detection of contacts (desired or non-desired ones). My talk we review our recent developments and views in tools in challenging this problem using only minimal sensor setting and also some larger view of the problem of inferring contact forces without using classical force sensing devices.
Title: "Contact detection for humanoids in close proximity interaction"
Abstract: Humanoid robots can serve as human partners in various close-contact situations. As real-use application perspectives appeared recently (domotics, large-scale manufacturing…), humanoids have great potential to be exploited as sophisticated assistive or cobotic systems. Their shape imparts them interesting properties in terms of integration, interaction with humans, empathy, and embodiment. As a key element in human-humanoid physical interaction is the detection of contacts (desired or non-desired ones). My talk we review our recent developments and views in tools in challenging this problem using only minimal sensor setting and also some larger view of the problem of inferring contact forces without using classical force sensing devices.
Dr. Kulić, Dana, Electrical and Computer Engineering, University of Waterloo, Canada.
Title: "Estimating, Modeling and Predicting Human Motion"
Abstract: Improved understanding and modeling of human movement can be used to teach robots to perform tasks, allow robots to safely and intuitively interact with humans, and to provide assessment and appropriate assistance to restore and facilitate movement. In this talk, I will describe tools for human motion measurement and analysis suitable for on-line applications that incorporate motion constraints. In the first part of the talk, a method for on-line pose estimation that exploits the geometry of the skeletal structure and motion constraints will be described. In the second part of the talk, approaches for motion modeling and reproduction will be introduced. The proposed method creates a generative model of the motion that can be used for motion segmentation and prediction. Results on a variety of datasets, including experiments in rehabilitation settings, will illustrate the proposed methods.
Title: "Estimating, Modeling and Predicting Human Motion"
Abstract: Improved understanding and modeling of human movement can be used to teach robots to perform tasks, allow robots to safely and intuitively interact with humans, and to provide assessment and appropriate assistance to restore and facilitate movement. In this talk, I will describe tools for human motion measurement and analysis suitable for on-line applications that incorporate motion constraints. In the first part of the talk, a method for on-line pose estimation that exploits the geometry of the skeletal structure and motion constraints will be described. In the second part of the talk, approaches for motion modeling and reproduction will be introduced. The proposed method creates a generative model of the motion that can be used for motion segmentation and prediction. Results on a variety of datasets, including experiments in rehabilitation settings, will illustrate the proposed methods.
Prof. Torras Genís, Carme, Institut de Robòtica i Informàtica industrial - UPC, Spain.
Title: "Dimensionality reduction in robot learning of impedance skills for cloth manipulation"
Abstract: Learning to manipulate clothing at the motion level is a complex endeavor. Uncertainties and force interactions arise that increase the difficulty of classical motion learning problems, characterized by an already high-dimensional parameter search space. In this context, dimensionality reduction techniques permit compressing motion descriptions and thus controlling the robot in a smaller state/action space, helping to learn cloth manipulation skills more efficiently.
Title: "Dimensionality reduction in robot learning of impedance skills for cloth manipulation"
Abstract: Learning to manipulate clothing at the motion level is a complex endeavor. Uncertainties and force interactions arise that increase the difficulty of classical motion learning problems, characterized by an already high-dimensional parameter search space. In this context, dimensionality reduction techniques permit compressing motion descriptions and thus controlling the robot in a smaller state/action space, helping to learn cloth manipulation skills more efficiently.
