Title: Bayesian Learning for Collaborative Source Seeking with Human-Robot Teams

Date: Tuesday, January 10, 2023

Time: 2:00PM – 4:00PM ET

Location: Teams Meeting (Meeting ID: 276 388 836 07  Passcode: fQDUZW)

Yingke Li

Robotics PhD Student

School of Electrical and Computer Engineering

Georgia Institute of Technology

Committee:

Dr. Fumin Zhang (Advisor) – School of Electrical and Computer Engineering, Georgia Institute of Technology

Dr. Enlu Zhou – School of Industrial and Systems Engineering, Georgia Institute of Technology

Dr. Matthieu Bloch – School of Electrical and Computer Engineering, Georgia Institute of Technology

Dr. Seth Hutchinson – School of Interactive Computing, Georgia Institute of Technology

Dr. Diyi Yang – Computer Science Department, Stanford University

Abstract: 

This proposal aims to devise a framework which enables shared autonomy for human-robot collaborative source seeking in turbulent environments, where the robots are tasked to simultaneously estimate the location of the source by sensor information and navigate towards the source, while in the meanwhile incorporating the superior situation awareness, logic, and problem-solving capability of their human partners. This framework empowers the robots with the full autonomy of SENSE, PLAN and ACT, as well as the versatility to smoothly adapt their autonomy level to achieve better team performance by INTERACTing with human. To overcome the multiple sources of uncertainty encountered that may thwart the success of collaborative source seeking in turbulent environments, this work resorts to a probabilistic perspective and a Bayesian framework is proposed to deal with different aspects of uncertainty. Specifically, Bayesian Consistent Data Fusion (DF) approach is utilized to incorporate the information from noisy measurements collected by heterogeneous sensors; Bayesian Risk-averse Model Predictive Control (MPC) is employed to warrant safe movement even in the presence of fluctuant dynamics caused by the turbulent flow; furthermore, Bayesian Mutual Theory of Mind (ToM) is exploited to infer the ambiguous intentions of human during the interaction.