CoRL 2020, Spotlight Talk 368: Time-Bounded Mission Planning in Time-Varying Domains with Semi-MDPs and Gaussian Processes

CoRL 2020

CoRL 2020, Spotlight Talk 368: Time-Bounded Mission Planning in Time-Varying Domains with Semi-MDPs and Gaussian Processes

Dec 16, 2020
|
35 views
Details
**Time-Bounded Mission Planning in Time-Varying Domains with Semi-MDPs and Gaussian Processes** Paul Duckworth (University of Oxford)*; Bruno Lacerda (University of Oxford); Nick Hawes (Oxford Robotics Institute) Publication: http://corlconf.github.io/paper_368/ **Abstract** Uncertain, time-varying dynamic environments are ubiquitous in real world robotics. We propose an online planning framework to address time-bounded missions under time-varying dynamics, where those dynamics affect the duration and outcome of actions. We pose such problems as semi-Markov decision processes, where actions have a duration distributed according to an a priori unknown time-varying function. Our approach maintains a belief over this function, and time is propagated through a discrete search tree that efficiently maintains a subset of reachable states. We show improved mission performance on a marine vehicle simulator acting under real-world spatio-temporal ocean currents, and demonstrate the ability to solve co-safe linear temporal logic problems, which are more complex than the reachability problems tackled in previous approaches.

Comments
loading...