Using Reinforcement Learning Routing Algorithm In WSNs

Wireless Sensor Networks(WSNs) is a self-organizing network which constitutes by a large number of sensors through wireless communication technology. The sensor nodes in the network can be extensively disposed to the detection area to monitor, to sensing and collecting information collaboratively. Sensor nodes cooperate in monitoring, sensing and information gathering(such as temperature, sound, vibration, pressure, movement, pollution). The collected information is transmitted to the receiver. WSNs application in military defense, environmental monitoring, disaster relief and commercial applications, and many other areas, including industrial monitoring and control, machine state detection, environment and animal monitoring, medical projects, home automation, traffic control, etc. The research of wireless sensor networks has been highly valued by academia and industry.Because of the weak ability of single node, traditional Internet routing protocol is not suitable for the wireless sensor networks. Design data routing algorithm is a challenging task since the requirement of save energy and reduce the processing complexity in wireless sensor network.This paper analyzes the main wireless sensor network routing protocol in recent years, according to the forwarding redundancy data problem in the propagation stage of directed diffusion protocol which is typical plane routing protocol, an optimal forwarding strategy(Hops Sense Directed Diffusion, HS-DD) based on hop count is proposed. This strategy reduce the number of nodes in the network to forward the message of Interest, and reduce the total number of messages of Interest. According to the single sensor information processing mechanism, combined with machine learning in reinforcement learning algorithm and the local routing information, an Energy Consumption Balance and Hop Less Adaptive Routing Algorithm which does not require complete topology information was proposed.(Energy Consumption Balance and Hop Less Adaptive Routing Algorithm, ECBHLA). Routing mechanism of ECBHLA is similar to the one of directional diffusion which is centered by data and request-driven, this mechanism by way of strengthening way to form the best path. ECBHLA information processing and forwarding mechanism using reinforcement learning algorithm commonly used algorithm in Q-learning. Qlearning algorithm can route using local information, after several routing, by analysing the routing information, it’s learn strategies to adapt to changes in the overall energy of the network nodes and the topology changes.Through the comparison of experiment analysis, ECBHLA get neighbor state information by feedback learning and control the way of forwarding data. Using that ECBHLA can choose the neighbor to which forward data need less hop and more energy. This property lead to effectively balance the overall energy of network and find less hop path to routing. Finally prolong the wireless sensor network life cycle.