Perception-Driven-Controlled UAV Interception and Collision Technology
WANG Zhibo1, HU Weijun1, MA Xianlong1, QUAN Jiale2, ZHOU Haoyu3
1. School of Astronautics, Northwestern Polytechnical University, Xi’an 710072, Shaanxi, China;
2. School of Aeronautics and Astronautics, Xi’an Jiaotong University, Xi’an 710049, Shaanxi, China;
3. Zhiyuan Research Institute, Hangzhou 310012, Zhejiang, China
Abstract:Unmanned Aerial Vehicles (UAVs) possess inherent advantages including high maneuverability, compact size, ease of modification, low-altitude slow flight capability, and adaptability to complex environments. These features enable tasks to be completed efficiently and gather information effectively. However, the same advantages also introduce security risks from unauthorized UAVs operating in restricted areas, which are compounded by the significant interception challenges posed by their high speed, low altitude, and highly maneuverable characteristics. To address these challenges, this study focused on the critical technical problems of autonomous decision-making and trajectory control for high-speed, precise UAV interception. A method for intercepting target UAVs using an interceptor UAV was proposed, whose core approach was an end-to-end deep reinforcement learning (DRL) network framework. This framework utilized the proximal policy optimization (PPO) algorithm to train a neural network controller that directly maps perceptual information to forces and torques controlling the quadrotor UAV. To optimize interception performance, a novel reward function based on reward shaping techniques was designed. This function enabled the interceptor UAV to achieve faster, smoother, and more precise interception trajectories. Experimental results demonstrate that the proposed method allows high-speed interception and achieves superior interception accuracy. Throughout interactions within the environment, the interceptor UAV demonstrates robust adaptive adjustment and real-time decision-making capabilities. This study verifies the effectiveness of the end-to-end DRL-based UAV interception solution. The method efficiently and precisely accomplishes high-speed interception tasks, while its end-to-end nature significantly reduces reliance on precise UAV dynamic models.
王志博, 呼卫军, 马先龙, 全家乐, 周皓宇. 感知驱动控制的无人机拦截碰撞技术[J]. 空天防御, 2025, 8(4): 78-84.
WANG Zhibo, HU Weijun, MA Xianlong, QUAN Jiale, ZHOU Haoyu. Perception-Driven-Controlled UAV Interception and Collision Technology. Air & Space Defense, 2025, 8(4): 78-84.