Minimizing Makespan with LaCAM* for Optimal Multi-Agent Path Finding

Omer Idgar, Dor Atzmon, Ariel Felner

Keywords: Multi-Agent Path Finding, Makespan, LaCAM*, A*

Abstract:

Multi-Agent Path Finding (MAPF) requires collision-free paths for
a set of agents. Optimal MAPF solutions often minimize one of
two cost functions: (1) sum-of-costs (SOC), which is the sum of the
costs of the paths, or (2) makespan (MKS), the maximum of these
costs. LaCAM* is a recent anytime MAPF solver, which eventually
converges to the optimal solution. However, LaCAM* has a considerably
slow convergence speed to the optimum and is considered a
scalable near-optimal solver. While this is true for SOC, in this paper,
we show that LaCAM* can quickly converge to the optimal MKS
solution. We explore LaCAM* and an improved version LaCAM*2
with three search methods: DFBnB, A*, and IDA*. We present their
superb performance for finding optimal MKS solutions, even with
thousands of agents.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 27