The Influence of Fixed and Moving NPC on Pedestrians’ Avoidance Behaviors: VR-Based Experiments

Abstract

Pedestrians have to take actions when crossing other pedestrians to avoid collisions. In this work, we focus on the differences of avoidance behaviors when a pedestrian crosses a moving and fixed intruder (NPC) in the virtual environment. The avoidance process is divided into three stages using the start avoidance point and maximum lateral deviation point. In moving NPC experiments, the distance from start avoidance point to the potential collision point (CP) first decreases and then increases as the intrusion angle increases. In standing NPC experiments, pedestrians start avoidance closer to the CP (average distance: 3.73m). In moving NPC experiments, the average maximum lateral offset distance (MLD) for the pedestrians to detour decreases with the intrusion angles decreases (Behind MLD ∈[1.09 m, 1.94 m], Front MLD ∈[1.13 m, 1.56 m]). In standing NPC experiments, the average MLD is 1.01m (left: 1.04m, right: 0.98m), which is the closest to the MLD of pedestrians at 180° intrusion angles. What’s more, at 30°, 60°, 90° and 120° intrusion angles, pedestrians avoiding behind the NPC require higher MLD than others avoiding in front of the NPC. Thus, more subjects prefer to avoid in front of the NPC under these conditions (88%, 86%, 78%, 69% of all). But the preference weakens and disappears at 150° and 180° intrusion angles due to the decrease of MLD. In standing NPC experiments, significant left-right preference is not found in pedestrians’ avoidance strategies (right: 46%, left: 54%). This article quantitatively analyses the difference between the influence of fixed and movement NPC on pedestrians’ avoidance strategies. The mechanism of pedestrian’s avoidance behavior is obtained by analyzing characteristic parameters, which is helpful to adjust pedestrian avoidance prediction models and design humanoid robots.