基于强化学习的多目标点航关联方法  

Multi-target Point-Track Association Method Based on Reinforcement Learning

作　　者：丁国胜[1] 蔡民杰 DING Guo-sheng;CAI Min-jie(Nanjing Research Institute of Electronics Technology,Nanjing 210039,China)

机构地区：[1]南京电子技术研究所,江苏南京210039

出　　处：《指挥控制与仿真》2022年第2期43-48,共6页Command Control & Simulation

摘　　要：针对密集杂波环境下的多目标点迹-航迹关联问题,以强化学习(Reinforcement Learning,RL)方法为基础,提出了一种基于Q学习的多目标点迹-航迹关联方法。首先,根据整个过程中目标的运动状态,建立马尔可夫决策过程(Markov Decision Process,MDP)模型。其次,利用各状态间的相关程度构成策略函数,选择准确的动作,并设定相应的奖励函数。最后,考虑杂波密集时虚假量测难以分辨,结合目标先验信息,增加了Q表再学习环节,进一步优化关联精度。仿真结果表明,在非机动和强机动两种环境下,该方法都能准确地关联到目标的量测,具有较好的点迹-航迹关联性能。Aiming at the problem of multi-target point-track association in dense clutter environment,based on the reinforcement learning(RL)method,a multi-target point-track association method based on Q-learning is proposed.First,according to the movement state of the target in the whole process,a Markov decision process(MDP)model is established.Secondly,the paper uses the degree of correlation between the states to form a strategy function,selects the correct action,and sets the corresponding reward function.Finally,considering that false measurements are difficult to distinguish when the clutter is dense,combined with the prior information of the target,the Q-meter re-learning link is added to further optimize the correlation accuracy.The simulation results show that in both non-maneuvering and strong maneuvering environments,the method in this paper can accurately correlate to the measurement of the target,and has a better point-track-track correlation performance.

关 键 词：多目标点迹-航迹关联 强化学习 MDP模型 策略函数 Q表再学习 

分 类 号：E917[军事]