Adaptive Anytime Multi-Agent Path Finding Using Bandit-Based Large Neighborhood Search

TL;DR

This paper introduces BALANCE, an adaptive multi-armed bandit approach for large neighborhood search in anytime multi-agent path finding, significantly improving solution quality without extensive prior tuning.

Contribution

Proposes BALANCE, a novel online learning method using multi-armed bandits to adapt neighborhood size and heuristics dynamically during MAPF optimization.

Findings

Achieves at least 50% cost improvement over state-of-the-art methods.

Thompson Sampling outperforms other bandit algorithms in this context.

Demonstrates effectiveness across multiple benchmark maps.

Abstract

Anytime multi-agent path finding (MAPF) is a promising approach to scalable path optimization in large-scale multi-agent systems. State-of-the-art anytime MAPF is based on Large Neighborhood Search (LNS), where a fast initial solution is iteratively optimized by destroying and repairing a fixed number of parts, i.e., the neighborhood, of the solution, using randomized destroy heuristics and prioritized planning. Despite their recent success in various MAPF instances, current LNS-based approaches lack exploration and flexibility due to greedy optimization with a fixed neighborhood size which can lead to low quality solutions in general. So far, these limitations have been addressed with extensive prior effort in tuning or offline machine learning beyond actual planning. In this paper, we focus on online learning in LNS and propose Bandit-based Adaptive LArge Neighborhood search Combined with Exploration (BALANCE). BALANCE uses a bi-level multi-armed bandit scheme to adapt the selection of destroy heuristics and neighborhood sizes on the fly during search. We evaluate BALANCE on multiple maps from the MAPF benchmark set and empirically demonstrate cost improvements of at least 50% compared to state-of-the-art anytime MAPF in large-scale scenarios. We find that Thompson Sampling performs particularly well compared to alternative multi-armed bandit algorithms.