Collaborative Strategy Learning Based Transmit Resource Management Scheme for Multiple Target Tracking in Multi-Cooperative Jamming Environment

This paper proposes a collaborative strategy learning based transmit resource management (CSLTRM) scheme for radar systems in a multi-cooperative jamming environment. Its key idea is to utilize the optimization technique to coordinate the transmit power and frequency in order to improve the target tracking accuracy. To combat multi-cooperative jamming, we first build an active MDP to sense the potential collaborative strategy within a Reinforcement Learning framework. Combining the reward and state transition function of the MDP, a weighted predictable Bayesian Cramér–Rao lower bound is constructed to evaluate the target tracking performance under the multi-cooperative jamming environment, and then the CSLTRM scheme is designed as a mixed-integer optimization model. In order to effectively obtain the joint transmit scheme, we develop a cyclic alternating minimization (CAM) method to solve the resulting optimization problem. Specifically, the CAM consists of two key steps: frequencies selection and power allocation. We provide a lower bound for the frequencies selection problem and thereby reduce the action space of the original optimization model. Then, the final transmit frequencies are further determined with the help of the cooperative information implicit in the anti-jamming model. After the transmit frequencies are given, the power allocation problem is solved by a convex optimization method. Simulation results also show that the proposed method can effectively reduce the impact of multi-cooperative jamming while improving target tracking accuracy with a limited power budget.