Multi-target dynamic hunting strategy based on improved K-means and auction algorithm

In this paper, a multi-target dynamic hunting strategy is proposed for the multi-agent system to hunt multiple dynamic targets, which can reasonably allocate the resources of the multi-agent system to complete hunting efficiently. The proposed strategy includes two parts: agent allocation and hunting control. In the agent allocation, the improved K-means algorithm is designed to divide the multi-agent system and dynamic targets into multiple independent single-target hunting subsystems. In the subsystem, the single-target hunting problem is decomposed into multiple subtasks and an auction algorithm is used to establish the correspondence between subtasks and agents. In hunting control, a controller is presented based on the backstepping method, which enables the agent to complete the subtasks. The command filter is introduced to address differential explosion. The stability of the controller is proved by the designed Lyapunov function. Simulation results show that the proposed strategy can reasonably allocate agents and enable the multi-agent system to hunt multiple dynamic targets.