On the Benefits of Robot Platooning for Navigating Crowded Environments

TL;DR

It is shown that for navigating passive and counter-flow crowds, the platooning strategy is less disruptive and more effective in dense crowds than the greedy strategy, whereas for navigating perpendicular-flow crowds, the greedy strategy outperforms the platooning strategy in either aspect.

Abstract

This paper studies how groups of robots can effectively navigate through a crowd of agents. It quantifies the performance of platooning and less constrained, greedy strategies, and the extent to which these strategies disrupt the crowd agents. Three scenarios are considered: (i) passive crowds, (ii) counter-flow crowds, and (iii) perpendicular-flow crowds. Through simulations consisting of up to 200 robots, we show that for navigating passive and counter-flow crowds, the platooning strategy is less disruptive and more effective in dense crowds than the greedy strategy, whereas for navigating perpendicular-flow crowds, the greedy strategy outperforms the platooning strategy in either aspect. Moreover, we propose an adaptive strategy that can switch between platooning and greedy behavioral states, and demonstrate that it combines the strengths of both strategies in all the scenarios considered.

Figures (7)

• Figure 1: Three crowded environments (not to scale). A group of robots (yellow disks) has to move from a start region ($\mathcal{W}_S$) to a goal region ($\mathcal{W}_G$) while traversing a region ($\mathcal{W}_C$) where they encounter a crowd of agents (gray disks). The solid black lines represent the boundary of the environment, while the dashed black lines represent a periodic boundary. (a) Crowd agents seek to remain stationary. (b--c) Crowd agents seek to move in the direction indicated by the white arrow. The dotted gray line further restricts the movements of crowd agents. In (b), any crowd agent crossing the dashed gray line into $\mathcal{W}_S$ is removed and a new crowd agent is added into $\mathcal{W}_C \cap \mathcal{W}_G$.
• Figure 2: Average crowd velocity in their desired direction. The solid green and dashed yellow lines show how the average crowd velocity decreases as the crowd densities for counter-flow and perpendicular-flow increase, respectively.
• Figure 3: Passive crowd navigation: (a) time taken for all robots to reach the goal region, and (b) number of comfort zone interceptions with the crowd.
• Figure 4: Counter-flow crowd navigation: (a) time taken for all robots to reach the goal region, and (b) number of comfort zone interceptions with the crowd.
• Figure 5: Perpendicular-flow crowd navigation: (a) time taken for all robots to reach the goal region, and (b) number of comfort zone interceptions with the crowd. The platoon (green) failed to reach the goal before the timeout of 900 s for densities of 0.2, 0.25, and 0.3 in 20%, 33%, and 67% of the trials, respectively.