Abstract: Aiming at the problem of radar main lobe suppression jamming in the combat scene of fixed wing UAV group shielding penetration platform, an intelligent trajectory planning method for UAV group based on reinforcement learning is proposed. In this paper, the two-stage suppression jamming problem of "approach" and "sustained jamming" is modeled as a Markov decision process, considering the complex conditions of UAV group long-range initiation and nose radiation orientation constraints. On this basis, a task allocation mechanism based on dubins distance is proposed, and an adaptive step size dual delay depth deterministic strategy gradient algorithm is designed to realize the integrated collaborative optimization of two-stage interference tasks. The simulation results show that the trajectory planning method in this paper can realize the UAV group from different positions, establish and maintain the main lobe suppression jamming based on the radar line of sight relationship, and effectively improve the jamming performance of UAV group under complex conditions.