Synthesizing Policies That Account For Human Execution Errors Caused By State Aliasing In Markov Decision Processes
Published in ICAPS Workshop on Explainable AI in Planning (XAIP 2021), 2021
Recommended citation: Gopalakrishnan, S., Verma, S., & Kambhampati, S. (2021). Synthesizing Policies That Account For Human Execution Errors Caused By State Aliasing In Markov Decision Processes. In ICAPS XAIP Workshop. https://arxiv.org/abs/2108.01777
Abstract
Humans assisting or executing sequential tasks frequently confuse perceptual states that appear visually identical despite having distinct underlying transition properties. We model this perceptual state aliasing within MDPs and synthesize assistive policies that avoid catastrophic error states.
Recommended citation: Gopalakrishnan, S., Verma, S., & Kambhampati, S. (2021). Synthesizing Policies That Account For Human Execution Errors Caused By State Aliasing In Markov Decision Processes. In ICAPS XAIP Workshop.