Authors:
Kentaro Sugiura
;
Mizuho Aoki
;
Kazuhide Kuroda
;
Hiroyuki Okuda
and
Tatsuya Suzuki
Affiliation:
Mechanical Systems Engineering, Nagoya University, Furo-cho, Chikusa, Nagoya, Aichi, Japan
Keyword(s):
Autonomous Mobile Robot (AMR), Pedestrian Behavior, Logistic Regression Model, Controllability.
Abstract:
Recently, a growing number of autonomous mobile robots (AMR) coexisting with humans are being introduced in many types of AMR-human shared space. Such AMR often needs to be navigated in narrow spaces while smoothly interacting with pedestrians. In such a situation, AMRs are highly recommended to estimate the pedestrian’s intentions and take appropriate action from the viewpoint of social acceptance. First, this paper presents new modeling and understanding of pedestrian behavior, particularly focusing on decision-making when they face an AMR at a close distance. Real-world experiments were conducted using a remote switch to directly record their decisions, and a mathematical decision model is made by using a logistic regression model. In the interaction between AMR and pedestrians, the AMR is expected to ‘implicitly control’ the interacting pedestrian by changing its own action. From this perspective, the influence of the AMR motion on the pedestrian’s decision is formally defined an
d calculated by using the controllability Gramian of the augmented AMR-pedestrian system model. A deep understanding of the influence of AMR action on pedestrian behavior will be beneficial to develop control policies for smooth AMR-pedestrian interactions.
(More)