Influence of person- and situation-specific characteristics on collision avoidance behavior in human locomotion.

Abstract

In everyday situations, pedestrians deploy successful strategies to avoid collisions with other persons crossing their paths. In this study, 2 experiments were conducted to investigate to what extent personal or situational characteristics affect role attribution and contribution to successful collision avoidance in human locomotion. Pairs of subjects walked at their natural speed from a start to a goal point. Walking paths were defined in such a way that subjects would collide halfway on their trajectory, if they did not actively avoid colliding by speed or path adjustments. In the first experiment, we investigated whether crossing order, path, and speed adjustments correlate with subject-specific parameters, such as gender, height, and personality traits. It is interesting that individuals' collision avoidance behavior was not correlated with any of these factors. In the second experiment, initial walking speed and heading were used to predict the crossing order. It was found that these 2 parameters are sufficient to estimate future role attribution with 95% confidence already 2.5 m before the crossing; that is, even before any collision avoidance behavior is initiated. In sum, this suggests that collision avoidance strategies in human locomotion are based on situational rather than on personal characteristics. These situational characteristics result in role attributions, which are highly predictable within and across pairs of pedestrians, whereby the role-dependent contribution of the pedestrian giving way is of greater relevance for successful collision avoidance. (PsycINFO Database Record

Minimum predicted distance: Applying a common metric to collision avoidance strategies between children and adult walkers

“…Collision avoidance strategy between two adult walkers is the result of mutually controlling their Minimal Predicted Distance (MPD) [7]. MPD accurately predicts the distance at which two walkers would collide if their speed and path trajectory were maintained at first sight of one another [7][8][9]. MPD calculation can be broken down into three successive stages: 1) observation phase, first sight of one another and MPD is low (i.e., a change in speed and trajectory is required in order to avoid a collision); 2) reactive phase, both individuals adapt their speed and trajectories to avoid a collision (i.e., MPD increases to 1.0m between the centre of the two young adult walkers [7]); and 3) regulation phase, the maintenance of MPD (i.e., the two walkers ensuring a collision does not occur).…”

“…Collision avoidance between two young adult walkers on a 90 degree collision course also involves role-dependent strategies. The decision of which walker crosses in front is mutually determined by the two walkers' path trajectories and velocities, which involves one walker increasing or maintaining their velocity to cross first while the other walker reduces their velocity and/or orientation to cross second [8][9]. The mutual interaction between two walkers is dependent on the visual information perceived by the two walkers allowing for successful collision avoidance behaviours [10].…”

Effective Human-Robot Collaboration in near symmetry collision scenarios

“…There have been several research works on collision avoidance between people during locomotor task. More specifically, previous works highlighted the collaborative nature of this task and that the relative collision avoidance contribution between people depends on crossing order [3] and that crossing order is determined at the start of the collision avoidance interaction between people [4].…”

“…Interpersonal coordination during collision avoidance has been shown to be mainly determined by situational factors such as properties of the environment, relative positions, speed and heading of people rather than their height, gender or personality [4]. However, planning collision avoidance motions solely based on a prediction of crossing order is insufficient to accommodate scenarios of uncertainty as in the absolute worst case of future collision, that happens in symmetric collision scenarios, choosing the most common crossing order is no better than chance because people would choose to cross first or last with equal probability.…”

“…Another approach, named Reciprocal Velocity Objects (RVO) [9] can also be used to replicate the behavior of crowds [10]. However, RVO was not modelled to naturally reproduce some behaviors of people during collision avoidance, for instance, people were shown to sometimes reverse their crossing order when avoiding collision [4] even though that entails longer time to avoid collision [1].…”