For my Planning, Search, and Reasoning Under Uncertainty course project, I and my partner did a replication study of a paper investigating the multi-agent path finding (MAPF) problem for car-like robots. This paper introduced a MAPF algorithm, car-like conflict-based search (CL-CBS), that incorporates cars’ kinematic and spatiotemporal constraints. This improves the algorithm’s transferability to real-world scenarios.

As part of the replication study, we performed additional experiments to investigate the algorithm’s performance on a variety of scenarios. The scenarios had varying map sizes, numbers of agents, Euclidean distances between start and goal states, and obstacle configurations.

These experiments helped to further understand the strengths/limitations of the algorithm.

The full report is below: