Showing 1–1 of 1 results for author: Zaro, A
-
Collision Detection for Multi-Robot Motion Planning with Efficient Quad-Tree Update and Skipping
Authors:
Abdel Zaro,
Ardalan Tajbakhsh,
Aaron M. Johnson
Abstract:
This paper presents a novel and efficient collision checking approach called Updating and Collision Check Skipping Quad-tree (USQ) for multi-robot motion planning. USQ extends the standard quad-tree data structure through a time-efficient update mechanism, which significantly reduces the total number of collision checks and the collision checking time. In addition, it handles transitions at the qu…
▽ More
This paper presents a novel and efficient collision checking approach called Updating and Collision Check Skipping Quad-tree (USQ) for multi-robot motion planning. USQ extends the standard quad-tree data structure through a time-efficient update mechanism, which significantly reduces the total number of collision checks and the collision checking time. In addition, it handles transitions at the quad-tree quadrant boundaries based on worst-case trajectories of agents. These extensions make quad-trees suitable for efficient collision checking in multi-robot motion planning of large robot teams. We evaluate the efficiency of USQ in comparison with Regenerating Quad-tree (RQ) from scratch at each timestep and naive pairwise collision checking across a variety of randomized environments. The results indicate that USQ significantly reduces the number of collision checks and the collision checking time compared to other baselines for different numbers of robots and map sizes. In a 50-robot experiment, USQ accurately detected all collisions, outperforming RQ which has longer run-times and/or misses up to 25% of collisions.
△ Less
Submitted 14 July, 2023;
originally announced July 2023.
