A game theoretic approach for antagonistic-task coordination of underwater autonomous robots in asymmetric threats scenarios

This work proposes a game theoretic approach to tackle the problem of multi-robot coordination in critical scenarios where communication is limited and the robots must accomplish different tasks simultaneously. An important application falls in underwater robotic framework where robots are used to protect a ship against asymmetric threats guaranteeing simultaneously the coverage of the area around the ship and the tracking of a possible intruder. The problem is modelled as a potential game for which novel learning protocols are introduced. Indeed, a general extension of pay-off based algorithms is herein proposed where the main difference with state-of-the-art protocols is that the trajectory optimization is considered instead of single action optimization. Moreover, the proposed T-algorithms, steer the robots toward Nash equilibria that will be shown to correspond to the accomplishment of different, possibly antagonistic, goals. Finally, performances of the algorithms, under different scenarios, have been evaluated in simulations.