Research On Waveform Selection Technology Of Cognitive Radar Seeker

In modern warfare,the environment is complex and ever-changing,and the traditional technical system of radar seeker is difficult to adapt to complex task requirements.Intelligent radar seeker has become an emerging direction of technological development.In response to the problem of waveform selection for maneuvering target tracking of cognitive radar seekers in cluttered environments,the paper constructs a radar seeker perception action cycle model,and proposes a waveform selection algorithm based on criterion functions,reinforcement learning,and multi-agent deep reinforcement learning,effectively improving the accuracy of seeker tracking and guidance.Firstly,the paper constructs a cognitive radar seeker perception action cycle,analyzes the radar seeker signal reception and processing model,target tracking and guidance model,and the relationship between radar waveform and measurement data error.Three radar waveform selection criteria functions are derived,providing theoretical support for subsequent waveform optimization and selection research.Secondly,in response to the waveform selection problem of cognitive radar seeker maneuvering target tracking in cluttered environments,the performance of a probabilistic data association algorithm assisted by measurement amplitude information was analyzed and derived,and the estimation of the covariance of IMMUKFPDA filter tracking error was achieved.On this basis,a radar seeker target tracking waveform selection algorithm based on three criteria functions is proposed,which effectively improves target tracking performance.The waveform selection algorithm based on the minimum mean square error criterion has the best comprehensive tracking performance.Compared to fixed waveform tracking,it can reduce distance tracking error by 74.94% and speed tracking error by 31.07%.At the same time,combining measurement amplitude information to assist tracking and the minimum mean square error criterion waveform selection method can reduce tracking distance error by 75.92%and speed error by 31.06% compared to fixed waveform tracking.Then,aiming at the problems of large amount of calculation and low self-learning ability in radar waveform selection of seeker,an intelligent waveform selection algorithm based on entropy reward Q-Learning and deep Q-network is proposed.By fitting Q-table with feedforward neural network,efficient optimal waveform strategy selection is realized.The simulation results show that the two proposed reinforcement learning algorithms can effectively improve the timeliness of waveform selection while improving the tracking accuracy of the seeker.Finally,in response to the waveform selection problem of cognitive radar seeker in collaborative tracking scenarios,a hierarchical data fusion global model for maneuvering target collaborative tracking was constructed.A cooperative multi-agent seeker collaborative tracking waveform selection algorithm based on independent deep Q-network was proposed,effectively improving the seeker’s collaborative tracking performance towards targets.The distance tracking error was reduced by 84.80%,and the speed tracking error was reduced by 54.87%.