Team formation is the problem of selecting a group of agents, where each agent has a set of skills; the aim is to accomplish a given mission (a set of tasks), where each task is made precise by a skill necessary for managing it. In a dynamic environment that o↵ers the possibility of losing agents during a mission, e.g., some agents break down, the ro-bustness of a team is crucial. In this paper, the focus is laid on the mission oriented robust multi-team formation problem. A formal framework is defined and two algorithms are provided to tackle this problem , namely, a complete and an approximate algorithm. In the experiments, these two algorithms are evaluated in RMASBench (a rescue multi-agent benchmarking platform used in the RoboCup Rescue Simulation League). We empirically show that (i) the approximate algorithm is more realistic for RMASBench compared to the complete algorithm and (ii) considering the robust mission multi-teams have a better control on the fire spread than the sophisticate solvers provided in RMASBench.