Cognitive Jammer Time Resource Scheduling With Imperfect Information via Fuzzy Q-Learning

Effective strategy generation of the jammer with inaccurate or undetermined information for combating the radar system is a challenging problem, and the relevant research is scarce in existing work. This article aims at exploring this promising topic and proposing a transmitting time slot scheduling strategy generation method for the cognitive jammer with imperfect time-matching information based on fuzzy reinforcement learning. First, the interaction between the jammer and environment is modeled as a Markov decision process (MDP) with uncertain state and continuous action, in which the goal of the jammer in interaction is to maximize accumulated discount pulse transmitting time slot jamming probability. Then, the uncertain pulse repetition intervals (PRIs) of the radar signal are inferred using the fuzzy inference system (FIS), where the inferred reward is evaluated by the correntropy of the inferred radar PRI and the previously measured one, thus presenting the fuzzy Q-learning based time scheduling (TSFQL) algorithm. Finally, numerical simulations in typical scenarios are performed to illustrate the effectiveness and superiority of the TSFQL algorithm over conventional methods.