关键词: 有限缓冲区, 多信道路由抗干扰, 分布式决策, 分层深度Q学习

Abstract: This paper investigates the multi-path routing anti-jamming problem in wireless self-organizing networks under multi-source concurrency. Due to limited node buffers and open wireless channels that are susceptible to jamming, critical nodes are prone to buffer overflow and link interruptions when undertaking multi-path data forwarding tasks, exacerbating network congestion. To this end, this article proposes a finite buffer-oriented multi-channel routing anti-jamming decision-making method to address the co-optimization of route planning, channel selection, and congestion control in dynamic jamming and high load environments. On this basis, the route planning and channel decision problem is modeled as a partially observable stochastic game model, and a multichannel routing anti-jamming decision algorithm based on hierarchical deep Q-learning is proposed: the upper layer network optimizes routing paths based on the neighbor buffer states and destination addresses to avoid congestion, the lower layer network combines the route decision-making with the spectral sensing results to dynamically avoid jamming channels, and a reward function is designed to achieve the optimal routing decision based on a fusion of the hop cost, the congestion penalty and the jamming avoidance. Through the design of reward function that combines hop cost, congestion penalty and jamming avoidance, the cooperative optimization of routing and anti-jamming channel access is realized. Simulation results show that compared with the comparison algorithm, the proposed algorithm improves the success rate of packet transmission by 15%, which effectively enhances the reliability and anti-jamming capability of wireless self-organized network in the jamming environment.

Key words: finite buffer, multi-channel routing anti-jamming, distributed decision making, hierarchical deep q-network