News

Doctoral Research Seminar on "Enhanced Robot Compliance, State Estimation, and Identification using Distributed Tactile Feedback: Leveraging Redundancy and Multimodality"


Doctoral Research Seminar on "Enhanced Robot Compliance, State Estimation, and Identification using Distributed Tactile Feedback: Leveraging Redundancy and Multimodality" by Quentin Leboutet, former ICS member.

Abstract:

In a rapidly evolving society, the prospect of short-term, large-scale deployment of a wide variety of autonomous robots, and in particular humanoid robots, for use in everyday life applications raises multiple questions and challenges, both from a societal and engineering perspective. Obviously, such robots must be endowed with high levels of adaptability in order to safely and effectively operate in a human environment.
In this regard, much research suggests that advanced sensory capabilities are essential to provide robots with an adequate level of situational awareness, allowing them to work side-by-side with humans or to execute enhanced whole-body manipulation tasks. Therefore, the implementation of an artificial sense of touch that can potentially equip any robot seems to be a very promising research path. 
Nevertheless, covering an entire robot with artificial skin remains problematic in practice due, on the one hand, to the substantial amount of information that can potentially be generated by this class of highly redundant systems and, on the other hand, to the fact that mounting an artificial skin on a robot is likely to significantly alters its dynamic properties. As a matter of fact, not only should the skin data be processed in real-time by robots with limited computational capabilities, but it should also be integrated into a more general control framework in order to generate suitable robot behaviors. The work of this thesis is organized around the following axes to provide a coherent response to several of these challenges.
First, the modalities of use of tactile information in the context of human-robot interaction are explored, through the development of whole-body tactile compliance control strategies.
Different applications of the skin for state estimation and inertial parameter identification purposes are then explored. In particular, the possibility of using the skin's inertial feedback to estimate robot joint motion derivatives is investigated, with possible applications to control and online parameter identification. This thesis then focuses on the development of an online identification strategy, typically for adaptive stabilization purposes of humanoid robots, leveraging the skin data to refine the underlying parameter identification process while limiting the excitation requirements, in a whole-body manipulation context.

Time: July 18th, 11:00-12:00 via zoom