Design And Optimization On Routing Technology In Wireless Sensor Networks

With its low deploayment cost, high measurement accurancy and long possibleworking time, Wireless Sensor Networks (referred as WSN) has been widely used incontrol and monitoring area. WSN is a distributed network system composed ofthousands of low-priced micro sensor nodes. Sensor nodes form a multi-hopself-organizing network through wireless communication, and transmit sensedinformation to the base station for further processing by some routing strategies. WSNcan be used in many area which including military, argricultural and industrial area.In these applications, data collection is the basic and the most important task forsensor nodes. In data collection task, sensor nodes are randomizedly or artificiallydeployed in the monitored area. After collecting required information, sensor nodestransmit them to some end nodes through routing paths and then end nodes send theseinfo via satellite or other means to a terminal instrument which can be directlyobserved by the testers. In the process of transmission, intermediate nodes also playthe role of routers, because they have to determine the next hop addresses to forwardthe information. As this self-organized routing strategy can greatly affect the networkperformance, it is always the one of most important and hot research topic.In general, two characteristics of WSN are important indicators to evaluate thepros and cons of routing strategy: the first one is the time of delay, which means howfast the info packets can be delivered to the base station. The second one is theprobably working time of WSN. Considering these two design criteria, we propose aMulti-Path Hierarchical Routing algorithm referred as MPHR. MPHR can take fulladvantage of the connection relationship between nodes in order to reduce thenetwork congestion and effectively balance the network load.Then aiming at application with uneven distribution of events, we propose anadaptive wireless sensor network routing algorithm MPHR-RL (MPHR usingPolicy-gradient Learning) based on MPHR which uses reinforcement learningtechnology. In this strategy, routing process is treated as a process that distributedintelligent nodes do reinforcement learning. Each sensor node is a stand-aloneintelligent node and determines the next-hop address by the parameterized probabilityand rewards. The routing scheme can make inert nodes take the place of busy-working nodes to transfer data, so that it can average nodes’ energyconsumption and prolong the whole network life time.At last, we do the simulation on these two routing strategy. And the results provethat our scheme can efficiently decentralize data transmission and prevent nodes fromearly dying.