Multi-Domain Resource Scheduling for Surveillance Radar Anti-Jamming based on Q-Learning

This paper proposes a multi-domain resources scheduling strategy based on Q-Learning for surveillance radar anti-jamming. As its core, the surveillance radar accomplishes detection tasks and avoids potential active jamming by intelligently selecting the radar transmit parameters of spatial (i.e., beam position), frequency, and energy (i.e., dwell time and transmit power) domains. The resources scheduling is formulated as a sequential decision problem in the context of unknown prior information about environments and enemy jammers. To describe this decision problem mathematically, we build a detailed Markov decision process (MDP) model and establish the corresponding reward function regarding the performance of the radar detection, low interception and the penalty after intercepted. A Q-Learning based solution is presented to find the optimized action strategy of the formulated model accordingly. Simulation results show that the proposed strategy can improve the radar detection performance while reducing the risk of interception by the enemy.