一种基于深度Q学习的移动Ad Hoc路由协议

A deep Q learning-based mobile Ad Hoc routing protocol

针对认知无线电-移动Ad Hoc网络,提出基于深度Q学习的移动Ad Hoc路由协议(DQRM)。DQRM路由先利用节点的剩余能量以及移动速度计算成本,再利用深度Q学习计算节点的Q值。在发现路由阶段时,源节点从邻居节点中选择具有最低Q值的节点传输RREQ包;利用单播方式传输RREQ包发现路由,降低了控制开销;目的节点收到RREQ包,沿着传输RREQ包的路径向源节点回复RREP包。仿真结果表明,提出的DQRM路由降低了控制开销,并提高了数据包传递率。

For cognitive radio-mobile Ad Hoc networks,a deep Q learning-based moblie Ad Hoc routing protocol(DQRM)is proposed in this paper.DQRM protocol uses the remaining energy and moving speed of nodes to calculate the cost,and then the deep Q is used to learn to calculate the Q value of nodes.During the discovery route phase,the source node selects the node with the lowest Q value from the neighbor nodes to transmit the RREQ packet.The unicast is used to transmit RREQ packet to find routing,which reduces the control cost.Once the destination node receives the RREQ packet,it replies the RREP packet to the source node along the path where the RREQ packet is transmitted.The simulation results show that the proposed DQRM routing reduces the control overhead and improves the packet transfer rate.