Learning to Improve Multi-Robot Hallway Navigation (2020)
Jin-Soo Park, Brian Tsang, Harel Yedidsion, Garrett Warnell, Daehyun Kyoung, and Peter Stone
As multi-robot applications become more prevalent, it becomes necessary to deve lop navigation systems which allow autonomous mobile robots to efficiently and safely pass each other in confined spaces. Existing navigation systems, such as the widely used ROS Navigation Stack, usually produce safe, collision free path s in static environments. However, these systems are not perfect, and when mult iple mobile robots simultaneously navigate in narrow spaces, collisions and tur narounds are not uncommon. Fine-tuning and enhancing such navigation stacks is not as simple as it looks since they are made up of multiple layers of code, an d there exists a tradeoff between optimizing for efficiency, i.e. minimizing ti me to destination (TTD) vs. optimizing for safety, i.e. minimizing collisions, with each objective leading to a different combination of parameter values. In this paper we develop a methodology to improve existing navigation stacks with regards to both objectives, without tuning their parameters, while preserving t heir inherent safety control properties. Our proposed approach is a decentraliz ed learning-based approach that is geared toward real world robotic deployment, by requiring little computing resources. It is agnostic of the underlying navi gation stack and can adapt to different types of environmental layouts (i.e., h allway structures).
In Proceedings of the 4th Conference on Robot Learning (CoRL), Virtual Conference, November 2020.

Peter Stone Faculty pstone [at] cs utexas edu
Garrett Warnell Research Scientist warnellg [at] cs utexas edu
Harel Yedidsion Postdoctoral Fellow harel [at] cs utexas edu