Improving Robot Predictability via Trajectory Optimization Using a Virtual Reality Testbed

Published: 27 Feb 2024, Last Modified: 04 Mar 2024VAM-HRI 2024 OralEveryoneRevisionsBibTeXCC BY 4.0
Keywords: hri, vr, trajectory optimization, predictability
TL;DR: We use VR as a testbed for our trajectory optimization approach to make robots more predictable to humans.
Abstract: The ability to predict where a robot will be next, or how it will navigate an area is critical to safe and effective human-robot collaboration and interaction. Due to information asymmetry, the path that a robot takes may be optimal, yet unpredictable to an observing human who does not have access to the same information. Unpredictability presents a safety risk to humans, and also makes interacting with robots more cognitively intensive and confusing than need be. In this work, we propose an algorithm that optimizes a robot's trajectory for predictable behavior, resulting in a robot that moves in a way that is more predictable to humans, balanced with what is optimal to the robot. To validate this approach, we propose two human-subjects experiments, one of which is conducted in virtual reality.
Submission Number: 12