k-Robust Multi-Agent Path Finding

This paper presents a planner that can solve the k-robust MAPF planner which is based on the Conflict-based search (CBS) (Sharon et al. 2015) MAPF solver. CBS does not explicitly search the n-agent state space. Instead, agents are associated with constraints of the form 〈ai, v, t〉, which would prohibit agent ai from occupying vertex v at time step t. A consistent path for agent ai is a path that satisfies all of ai’s constraints, and a consistent plan is a plan composed only of consistent paths. Note that a consistent plan can be invalid if it contains conflicts despite each path satisfying the individual agent constraints. CBS works by searching a constraint tree (CT) for a set of constraints such that a consistent plan w.r.t. this set of constraints is optimal. The CT is a binary tree, in which each node N contains: (1) a set of constraints imposed on the agents (N.constraints), (2) a single plan (N.π) consistent with these constraints, and (3) the cost ofN.π (N.cost).The root of the CT contains an empty set of constraints (thus, every plan is consistent with the root). Each vertex N has two successors, which are generated by first finding a conflict