Collision-Free Navigation of Robotic Swarms using Preferred Velocity Metrics and Time Prioritized Avoidance

David Rubel*, R. Michael Butts*, Stephen Jacobs*, Yu Gu, Ali Baheri, and Guilherme Pereira
Department of Mechanical & Aerospace Engineering, West Virginia University, Morgantown, WV 26506-6045

Presentation No.: 66

Assigned Category (Presentation Format): Engineering (Poster Presentations)

Student’s Major: Mathematics and Computer Science Dual Degree

Multi-agent collision avoidance requires a delicate balancing act between avoiding other agents and making progress towards a goal. Existing methods can already guarantee collision free trajectories, but their formulations often cause unnecessary stagnation when the number of agents increases. Furthermore, the metrics used to compare possible trajectories with the desired one are usually slow at making progress toward the goal. This study presents a novel approach for collision avoidance with a non-Euclidean metric for determining optimal velocities and dynamic restrictions that safely expand the possible velocities an agent can take. More specifically, the metric will heavily disincentivize any slowing or speeding up in the preferred direction of motion to maintain progress toward the goal, and the restrictions that prevent a collision will increase as the time until that collision decreases. Based on randomized waypoint navigation simulations, block paired data shows a 17% average improvement in time efficiency compared to the Optimal Reciprocal Collision Avoidance (ORCA) method, a widely accepted collision avoidance approach.

Funding: NSF, Award # 1851815

Program/mechanism supporting research/creative efforts: the WVU Robotics REU (Yu Gu & Jason Gross)