关键词: 任务驱动, 动态权重, 路由算法, 拓扑子网, 机器学习

Abstract:

At present, the path selection strategies of existing Ad Hoc network routing protocols have obvious deficiencies in the intelligent optimization of task classification and network state coordination. They are difficult to meet the growing demand for high-reliability and low-latency data transmission driven by strong node mobility and diverse task requirements in complex application scenarios. An event-driven adaptive routing protocol for Ad Hoc networks is proposed, aiming to address the communication challenges caused by dynamic environment changes and diverse task requirements in Ad Hoc networks. The protocol enables data transmission by planning multiple topological subnets for different tasks and introduces a dynamic weight adjustment mechonism. The Transformer-based traffic prediction algorithm is embedded into the dynamic routing adjustment module, which provides traffic prediction and path optimization support for the routing algorithms of three types of tasks （i.e., emergency, regular, and low-priority tasks）. This optimizes communication performance in complex network scenarios and realizes adoptive adjustment of routing strategies, ensuring real-time and reliable data transmission. Experimental results demonstrate that, compared to several traditional baseline routing protocols, the event-driven adaptive routing protocol improves the transmission efficiency of high-priority tasks by an average of 32.9% without significant loss in overall system performance. Additionally, it performs better in complex network scenarios compared to several traditional base line routing protocols, effectively meeting the communication needs of various task types.

Key words: event-driven, dynamic weight, routing algorithm, topological subnet, machine learning