P-ARC speeds up multi-robot planning 4x with parallel processing

A team from the University of Illinois at Urbana-Champaign has published P-ARC (Parallel Adaptive Robot Coordination), a parallelized version of the ARC algorithm designed to accelerate multi-robot motion planning (MRMP). The core insight is that ARC decomposes MRMP into subproblems that can be solved independently, and P-ARC exploits this by running initial individual solutions, conflict detection, and conflict resolution in parallel across multiple CPU cores. Additionally, the researchers introduce OR-P-ARC, a hybrid that combines the multistart strategy of OR-parallelism with P-ARC's pipeline parallelism.

In experiments with up to 128 robots in 2D mobile and planar manipulator scenarios, P-ARC demonstrated planning time speedups approaching 4x over the sequential version, specifically when deploying 16 CPU cores for large Panda multi-manipulator teams in real-world inspired environments. The paper (arXiv:2606.27625) shows how parallelizing each of the three main ARC stages can dramatically reduce wall-clock time without sacrificing solution quality, making multi-robot coordination more practical for industrial settings like warehouse automation or assembly lines.