Compilation-based techniques represent a significant stream of solvers for multi-agent path finding (MAPF) due to their modularity and adaptability to non-standard problem variants. In standard MAPF, agents navigate from their initial positions to given individual goal positions without collision. However, UA-MAPF introduces a variant where some agents maintain the same initial positions and goals, while others are unassigned, having only initial positions and no goals. Although unassigned agents do not need to reach a specific target, they must move out of the way of standard agents when necessary, presenting a unique challenge. This paper demonstrates that UA-MAPF can be formulated using recent compilation-based techniques for MAPF by framing the problem as Boolean satisfiability, specifically adapting SMT-CBS and NRF-SAT, which are recent solvers based on counterexample guided abstraction refinement and non-refined abstractions.
Blogger's Review: This paper investigates the UA-MAPF problem, proposing a novel approach to handle complex agent interactions, particularly in the absence of specific goals. This is crucial for enhancing the flexibility and adaptability of agent systems, and merits deeper exploration in practical applications.