基于空−海−潜跨域无人平台协同的海上目标探测追踪策略OA
Maritime Target Detection and Tracking Strategy Based on the Collaboration of Air-Sea-Submarine Cross-domain Unmanned Platform
提出一种基于空−海−潜跨域无人平台协同的海上目标探测追踪策略.首先,构建无人机−水面艇−潜器协同的海上跨域无人系统.然后,针对海上目标的高机动性以及无人平台自身约束,采用测度理论解析无人机−水面艇−潜器最佳探测编队队形,实现目标探测概率最大化;当探测到目标后,设计基于逆强化学习的无人机−水面艇−潜器编队控制器,实现障碍物环境下水面/水下目标的可靠有效追踪.最后,通过仿真与实验验证了所提方法的有效性.结果表明,所提探测模式可以实现有限时间内移动目标探测概率最大化,同时所提逆强化学习编队控制器可以在保持队形稳定的基础上,结合动态避障策略,实现复杂环境下跨域无人平台安全协同追踪.
This paper proposes a maritime target detection and tracking strategy based on the collaboration of air-sea-submarine cross-domain unmanned platform.Firstly,a maritime cross-domain unmanned system that integ-rates unmanned aerial vehicle(UAV),unmanned surface vessel(USV),and autonomous underwater vehicle(AUV)is constructed.Then,the optimal detection formation of UAV-USV-AUV is analyzed using measure theory to max-imize target detection probability,accounting for the high maneuverability of maritime targets and the constraints of unmanned platform;After the target is detected,an inverse reinforcement learning-based controller is designed for the formation of UAV-USV-AUV to achieve reliable and effective tracking of surface/underwater targets in obstacle-prone environments.Finally,simulations and experiments are conducted to validate the effectiveness of the proposed method.The results show that the proposed detection mode can maximize the probability of detecting mo-bile targets within a finite time,and the proposed inverse reinforcement learning formation controller can achieve secure and collaborative tracking of cross-domain unmanned platform in complex environments by combining dy-namic obstacle avoidance strategies while maintaining formation stability.
田泽兴;闫敬;高麒媛;杨晛;关新平
燕山大学电气工程学院 秦皇岛 066004||智能控制与神经信息处理教育部重点实验室 秦皇岛 066004燕山大学电气工程学院 秦皇岛 066004||智能控制与神经信息处理教育部重点实验室 秦皇岛 066004燕山大学电气工程学院 秦皇岛 066004||智能控制与神经信息处理教育部重点实验室 秦皇岛 066004燕山大学电气工程学院 秦皇岛 066004||智能控制与神经信息处理教育部重点实验室 秦皇岛 066004上海交通大学自动化与感知学院 上海 200240
探测追踪跨域无人平台避障逆强化学习
detectiontrackingcross-domain unmanned platformobstacle avoidanceinverse reinforcement learn-ing
《自动化学报》 2026 (2)
349-362,14
国家自然科学基金(62222314,U25A20472),河北省燕赵青年科学家项目(F2024203047),河北省自然科学基金(F2022203001,F2024203072,F2025501051),河北省教育厅基金(JCZX2025027)资助Supported by National Natural Science Foundation of China(62222314,U25A20472),Yanzhao Young Scientist Project of Hebei Province(F2024203047),Natural Science Foundation of Hebei Province(F2022203001,F2024203072,F2025501051),and Education Department Foundation of Hebei Province(JCZX2025027)
评论