Ariel Felner

Online MAPF extends the classical Multi-Agent Path Find ingproblem (MAPF) by considering a more realistic problem in which new agents may appear over time. As online solvers arenot aware of which agents will join in the future, the notion of snapshot-optimal was defined, where only current knowledge is considered. In this paper, we perform an extensive com-parison between oracle-optimal solutions (where the solver is preinformed of future agents), snapshot-optimal solutions, and suboptimal solutions obtained by prioritised planning