Rapid jamming recognition based on Q-learning sampling under resource constraints

In the context of limited resources, jamming sampling poses a bottleneck in improving jamming cognition performance. Existing approaches to jamming cognition predominantly focus on jamming identification methods, overlooking the crucial prelude of jamming data collection. This paper formulates the problem of resource-constrained jamming cognition as a sampling resource allocation issue and introduces a Q-learning-based adaptive jamming sampling method. The process of selecting sampling sub-bands under resource constraints is modeled as a Markov Decision Process (MDP), and a reward function is designed to incorporate the jamming recognition results obtained through a deep convolutional neural network into the sampling loop. Simulation results demonstrate that, compared to existing periodic sampling methods, the proposed algorithm significantly enhances the efficiency of effective sampling, achieving rapid cognition of jamming, whether the jamming sub-bands are fixed or variable.