Skin-Embodied Intelligence for Human-Robot Interaction
Dates: | 19 February 2024 |
Times: | 13:30 - 15:00 |
What is it: | Seminar |
Organiser: | Department of Electrical and Electronic Engineering |
Who is it for: | University staff, Current University students |
Speaker: | Prof Vladimir Lumelsky |
|
Welcome to this IEEE distinguished lecture
The ability by a robot to work near humans or at a faraway planet under “soft” human control akin to a supervisor-instructed human workers, rather than rigid pre-programmed instructions, opens a myriad of potential uses. Examples include robots preparing the Mars surface for human arrival; robot assembly of large space telescopes; robot helpers for the elderly; robot search and disposal of war mines. Handling uncertainty and guaranteeing robot safety become major issues, calling for way higher than in today’s robotics for mass car production and the like. Accordingly, “uncertainty-able” robots are still exceedingly limited and rare. Challenges appear both on the robotics and the human sides: robots have hard time adjusting to an unstructured/dynamic environment, whereas human cognition has serious limits in designing tasks that require orientation and motion in 2D and 3D space. Not surprisingly, human texts-based experience encapsulated in AI systems like ChatGPT are of little use in such tasks. The way out of this impasse is to supply the robot with sensing-based intelligence whereby real-time sensing information is fed into geometry and topology-based tools capable of reasoning about space and motion. An important case here is the skin-embodied intelligence - an ability widely spread in nature, including humans, to plan one’s motion in space by reasoning about surrounding objects based on sensing at one’s whole body. This calls for new reasoning algorithms and new hardware - whole-body sensing, a skin covering the robot’s body akin to the human skin. Such systems possess interesting, even unexpected, properties: heavy powerful robots become inherently safe; human operators can move them fast, with “natural” speeds; robot motion strategies exceed human spatial reasoning skills; there appears natural synergy of human-robot teams with arbitrary mix of supervised and unsupervised robot operation. We will review the mathematical, algorithmic, hardware (materials, electronics, computing) and cognitive science issues involved in realizing such systems.
Speaker
Prof Vladimir Lumelsky
Organisation: University of Wisconsin-Madison
Biography: Prof Zhongdong Wang (FIEEE, FIET, FCSEE) is a Professor of Electrical Energy System Infrastructure at Energy and Power Division of the Department of Electrical and Electronic Engineering at The University of Manchester. Her current research interests include condition monitoring techniques, modelling techniques and digital twin development for transformers and circuit breakers, environmentally friendly power equipment, and asset management.
Travel and Contact Information
Find event
Engineering A_3A.057 M&T
Engineering Building A
Booth St E