Analysis and Modeling of Human Behavioral Strategies in Collision Avoidance with Agents Employing Active and Passive Strategies

Abstract

This research aims to develop robots that can switch between passive strategies, which involve observing and reacting to a partner’s behavior, and active strategies, where they take the initiative to indicate their course of action, similar to humans. For instance, when passing someone on the street, a person might actively decide to move to the left or right in advance to avoid a collision. In this study, we conducted experiments in a grid environment where subjects encountered computer-controlled agents em-(breakpoint)ploying both passive and active strategies. The findings confirmed that the subjects’ behavioral strategies varied according to the strategies of the computer agents, providing significant insights into human-robot interaction. Additionally, the subjects’ behaviors showed differences compared to the strategies learned by the computer agents, further enriching our understanding of this complex field. These results suggest that even simple cooperative behaviors can be divided into finer cooperative tasks, and the consistency perceived in these behaviors varies among individuals. We propose a hierarchical model for strategy decision-making.