Temporal Logic Imitation: Learning Plan-Satisficing Motion Policies from DemonstrationsDownload PDF

16 Jun 2022, 10:45 (modified: 16 Nov 2022, 07:48)CoRL 2022 OralReaders: Everyone
Student First Author: yes
Keywords: Learning from Demonstration, Dynamical Systems, Formal Methods, Linear Temporal Logic, Certifiable Imitation Learning
TL;DR: Combining the task-level reactivity of LTL and the motion-level reactivity of DS, we arrive at an imitation learning system able to robustly perform various multi-step tasks under arbitrary perturbations given only a small number of demonstrations.
Abstract: Learning from demonstration (LfD) has successfully solved tasks featuring a long time horizon. However, when the problem complexity also includes human-in-the-loop perturbations, state-of-the-art approaches do not guarantee the successful reproduction of a task. In this work, we identify the roots of this challenge as the failure of a learned continuous policy to satisfy the discrete plan implicit in the demonstration. By utilizing modes (rather than subgoals) as the discrete abstraction and motion policies with both mode invariance and goal reachability properties, we prove our learned continuous policy can simulate any discrete plan specified by a linear temporal logic (LTL) formula. Consequently, an imitator is robust to both task- and motion-level perturbations and guaranteed to achieve task success.
Supplementary Material: zip
Website: https://sites.google.com/view/ltl-ds
21 Replies