歩行者の迷いを低減する移動ロボットのモデル予測型速度計画

抄録

Recently, the demand of Autonomous Mobile Robots (AMRs) has increased. However, the flexible navigation capabilities of AMRs can lead to unpredictable behavior from a pedestrian's perspective, potentially causing anxiety and hesitation during human-robot interactions. This paper presents a novel model predictive behavior planning method aimed at reducing pedestrian hesitation, specifically in crossing scenarios between an AMR and a pedestrian. A "decision entropy" metric to quantify pedestrian hesitation is defined and exploited in the model predictive control (MPC) framework. Real-world experiments involving an AMR and a pedestrian subjects are conducted under various crossing conditions. The effectiveness of the proposed method is evaluated through post-interaction questionnaires assessing clarity, safety, cooperativeness, and smoothness of robot behavior. Results from 384 crossing scenarios with six participants demonstrated that incorporating decision entropy reduction in the control objective significantly improved the perceived clarity of robot intentions and reduced pedestrian hesitation.