An optimal POMDP-based anti-jamming policy for cognitive radar

Abstract

This paper considers the radar resource management problem in the cognitive radar for tracking multiple moving targets in the presence of jammers. Since the radar is not able to obtain the accurate information of the moving targets due to the jamming, we formulate this problem as a hybrid partially observed Markov decision process-based (HPOMDP-based) game model to exploit the statistical characteristics of the moving targets and the competition between the jammers and the radar. In contrast to the existing methods of radar resource management, this work considers two competitive decision makers in the environment of anti-jamming radar, and deduces a symmetry expression of the performance gradients of the jammers and the radar. A low complexity joint optimization algorithm is proposed to find the optimal anti-jamming policy of the radar. Numerical results show that the proposed algorithm can obtain accurate detection results in the presence of jammers.