A novel multi-agent distributed cognitive frequency-selection model

To address the challenge of ensuring reliable communication for multi-agent systems operating in highly dynamic and complex electromagnetic environments, this paper proposes a novel distributed cognitive frequency-selection model. In this model, agents dynamically adjust their environmental sensing strategies based on access states, eliminating the need for a dedicated sensing channel. They perceive the protocol state transitions of neighboring nodes and the characteristics of spectrum and time-slot resources in real time, and thereby select optimal spectrum resources for communication. Specifically, the environmental cognition model is integrated with a distributed, partially observable Markov decision process to form the distributed cognitive frequency selection framework, and a corresponding algorithm is then designed based on the automatic repeat request protocol. By perceiving neighbor node information and extracting feature knowledge, the algorithm effectively extends the perception range to the logical domain of one-hop neighbors, thereby breaking through the limitations of traditional physical sensing, which is confined to single nodes. This approach not only significantly enhances the scope, timeliness, and accuracy of perception but also enables agents to rapidly avoid interference. Theoretical analysis and simulations demonstrate that in highly interfered and fast-varying scenarios, the proposed algorithm improves the packet delivery ratio by 10%∼30% compared to typical algorithms, outperforming existing spectrum sensing-based agent models. Without modifying the existing protocol designs, the algorithm maintains compatibility with current protocol architectures and spectrum sensing methods, allowing for seamless integration into heterogeneous networks for applications such as dynamic spectrum sharing and anti-jamming communications. This study provides new insights for creating reliable multi-agent communication in dynamic electromagnetic environments.