面向移动汇聚节点的无线传感器网络自适应路由方法

Adaptive Routing for Wireless Sensor Network with a Mobile Sink Using Reinforcement Learning

针对无线传感器网络面向移动汇聚节点的自适应路由问题,为实现路由过程中对节点能量以及计算、存储、通信资源的优化利用,并对数据传输时延和投递率等服务质量进行优化,提出一种基于强化学习的自适应路由方法,设计综合的奖赏函数以实现对能量、时延和投递率等多个指标的综合优化;从报文结构、路由初始化、路径选择等方面对路由协议进行详细设计,采用汇聚节点声明以及周期性洪泛机制加速收敛速度,从而支持汇聚节点的快速移动;理论分析表明基于强化学习的路由方法具备收敛快、协议开销低以及存储计算需求小等特点,能够适用于能量和资源受限的传感器节点;在仿真平台中通过性能评估和对比分析验证了所述自适应路由算法的可行性和优越性。

Many applications of wireless sensor network use a mobile sink to collect data detected by network nodes.To make routing approach in such scenarios better support sink mobility and be adaptive to varying network conditions,an adaptive routing approach based on reinforcement learning was proposed.A comprehensive reward function concerning mean hop count,residual energy and link quality was designed for learning to achieve a trade-off among multiple metrics of the network performance.Sink announcement was used to guarantee the convergence speed for obtaining an optimal route.Theoretical analysis shows that the approach is feasible for resources constrained sensor nodes due to its low overhead,small storage and computation requirements.Simulation results show the feasibility and superiority of the proposed approach.