USV Swarm Defense Optimization for Island Protection Based on Enhanced IDQ

With the development of unmanned technology, Unmanned Surface Vessel (USV) swarms can effectively safeguard maritime security, with island protection being a key aspect. It requires friendly USVs to develop effective strategies to destroy enemy USVs before they land on friendly islands. Therefore, to address the problem of generating swarm protection strategies in island protection scenario, this paper proposes a USV swarm defense strategy optimization method based on enhanced Independent Deep Q-learning (IDQ). The contributions of this paper are as follows: First, this problem is formulated as the Markov Decision Process considering the multiple islands protection scenario. Besides the traditional USV movement constraint, this model additionally considers the sector attack of the USVs under the multi-obstacle environment. Second, a primary-team mixed reward function is designed to accelerate the defense strategy learning process. The reward function is consisted by the primary target reward and team reward, while the primary target is selected based on the threaten level of the enemy USVs. Finally, an enhanced IDQ method is proposed to adapt to the varying number of enemy USVs. By incorporating the K-Means clustering, the enhanced IDQ focus on the aggregated behavior of the enemy cluster which avoids requiring all the status of the enemy USVs. Simulation results show that this method can effectively destroy enemy USV swarms and adapt to changes in enemy numbers. Comparative analysis demonstrates the superiority of the improved algorithm.