Trajectory Entropies for Autonomous Navigation and Active Sensing
Speaker
Girish Nair
Professor at the University of Melbourne.
Title
"Trajectory Entropies for Autonomous Navigation and Active Sensing"
Abstract
We discuss joint entropy-based measures of trajectory uncertainty and communication complexity for autonomous navigation and active sensing. These measures capture the minimum internal cost, in bits, of storing/communicating state trajectory beliefs and sensor observation sequences. In robotics, whole-of-trajectory entropies have previously been dismissed as intractable to optimise due to the nonlinear joint entropy functional acting on the entire trajectory. We show that, surprisingly, these entropies can be put into convenient stage-additive forms that enable optimisation using standard techniques, leading to principled trade-offs between exploitation and exploration. In simulations, trajectories reminiscent of the motion of some animals are produced. This is joint work with Dr Timothy L. Molloy (ANU) and Dr Changrong Liu.
References:
https://doi.org/10.1109/TAC.2023.3250159
About Speaker
Girish N. Nair is a Professor in the Department of Electrical and Electronic Engineering at The University of Melbourne. His research focuses on the interplay between estimation, control and information theory, in both stochastic and nonstochastic settings. He has received several prizes, including the 2014 George S. Axelby Outstanding Paper Award from the IEEE Control Systems Society and a 2006 SIAM Outstanding Paper Prize. He is a Fellow of the IEEE and was an Australian Research Council Future Fellow (2015 - 2019).
From 2019 - 2024 he was the lead Australian investigator for the Australia-US Multidisciplinary University Research Initiative on Neuro-Autonomy. From 2017 - 2022 he served as Deputy Head of Department (Research & Research Training), and from 2022 - 2024 he led the Control & Signal Processing research group in the Department. He is the General Chair of the 67th IEEE Conference on Decision and Control, to be held in Sydney, Australia in 2028, and the inaugural chair of the Victoria/New South Wales Joint Chapter of the IEEE Information Theory Society.