QMIX: Monotone Valued Function Decomposition Algorithm for Switching Formation-Contrianment Control with Local Information

In this paper, the formation-containment control problem of multiple unmanned aerial vehicles(UAVs) is studied. Note that the real systems are nonlinear and dynamic with uncertainty and hence their models are difficult to establish, such that it is difficult to adapt to some complex application scenarios combined with obstacle avoidance and navigation tasks. To solve this problem, a formation-containment control strategy based on a monotone valued function decomposition multi-agent reinforcement learning algorithm(QMIX) is designed, it only requires local neighboring information. In this paper, multiple UAVs can adapt to the complex and changeable situation in the battlefield and timely respond to communication failure. In simulations, it is shown that the proposed method is effective in the case of arbitrary topological shapes in dynamic scenarios, and owns the quick switching performance.