Intelligent collaborative control of UAV swarms with collision avoidance safety constraints

An intelligent cooperative control method for Unmanned Aerial Vehicle (UAV) swarms combining Deep Reinforcement Learning (DRL) and the collision avoidance strategy is proposed. First, the problem of UAV swarm control is described, and a motion model of the UAV is established. Secondly, under the framework of deep reinforcement learning, the training mechanism of the UAV swarm control strategy is established. By analyzing the observation space and action space of the UAV, the network structure of the UAV swarm control strategy is designed. The reward function is designed with the safety, tightness and consistency of the UAV swarm as the main goal. To solve the problem of lack of safety guarantee in the DRL method, the collision avoidance strategy of the UAV swarm is designed at the same time, and the collision avoidance commands can be calculated according to the relative motion states between UAVs and between UAVs and threat areas to avoid UAVs collision. The simulation results show that the proposed control method can effectively avoid the risk of collision between UAVs, guarantee the safety of the UAV swarm, and make the UAV swarm more compact.