Research Of Coverage Control Based On Reinforcement Learning In Wireless Sensor Networks

As the fundamental issues in Wireless Sensor Networks(WSNs), the network coveragereflects the "perceived" quality of service of WSNs, and its performance directly determinesusing result and the service experience of the networks. Moreover, the nodes in the networkshave low power energy and are difficult to replenish energy, so WSNs have the serious energyconstraint problem. Therefore, the network coverage control should balance and reduce theenergy consumption with ensuring the network coverage quality, which can achieve thepurposes of optimizing the network coverage. Meanwhile, as the new machine learningmechanism, reinforcement learning is just used to address the problem that an autonomousagent that can sense and act in its environment learns to choose the optimal actions controlstrategy to achieve its goals through the information interaction with the surroundingenvironment, the rewards and punishments mechanism of its selecting actions. Accordingly,based on reinforcement learning, it is the focus or this paper to designing reasonable andeffective network coverage control method and node scheduling strategy, as well as studyingthe coverage optimization algorithm considering the performance of both energy efficiencyand network coverage.This paper systematic analyses and summarizes the coverage problem in WSNs; basedon deeply studying the focus issue in current network coverage control algorithm, aiming atthe problems existed in coverage control algorithm based on reinforcement learning and otherexcellent coverage control strategies, this research presents corresponding measures. Thespecific work is summarized as follows:(1) This paper studies the characteristics of WSNs and summarize the classic coveragecontrol algorithm in both omni-directional and directional sensor networks; it analyses theoptimal control mechanism base on reinforcement learning and learning automata(LA), and itsummarized research of network coverage control based on the reinforcement learning inWSNs. Aiming at the key problems existed in network coverage and shortcomings in currentresearch of the coverage algorithm, based on the learning automata which is in the frameworkof reinforcement learning, this paper presents several coverage control algorithm with thecorresponding solution, and it also analyzes and validates the feasibility of the proposedalgorithms.(2) It has been the focus of the research on clustering protocols in wireless sensornetworks for cluster heads selection optimization and the energy load balancing among allsensor nodes to extend the network lifetime. Aiming at the random distribution of nodes inWSNs, basing on ICLA algorithm which adopts the learning automata, an energy balancedunequal clustering algorithm with the node density is proposed and evaluated in this paper. Inthe cluster head election phase, overall considering the residual energy and the node density,and moreover, adopting the LA for information exchange with the surrounding environment,it can choose relatively better cluster heads. According to the distance between cluster headsand the base station and the node density, it forms unequal clusters to balance energy load ofintra-and inter-clusters in different positions and node density degrees of networks. Thealgorithm adopts an evaluation function of neighbor cluster heads, which considers the energy of cluster head, node density in cluster and distance from each cluster head to the base station.so it can choose the transit cluster heads using greedy algorithm for multi-hop transmission.Meanwhile, it adopts the node scheduling scheme in cluster based on learning automata toreduce the number of active nodes in cluster and the network coverage redundancy.Simulation results show that it can choose relatively more reasonable cluster heads, efficientlybalance the energy load of all nodes and significantly prolong the network lifetime.(3) It has been the key problem of the research on coverage control protocols indirectional sensor networks for optimizing node sensing direction adjustment and reducingnetwork coverage redundancy to achieve long-term efficient network coverage. Based on thedeeply research on the problems of both network coverage-enhancement algorithm based onvirtual potential field and node scheduling strategies for redundant nodes, this paper presentsthe coverage control protocol based on both virtual potential field and learning automata. Inthe aspect of network coverage enhancement, it brings the improved virtual force model basedon both distance and overlapping sensing rate; considering the effect of virtual centripetalforce and tangential force on the perceived angle adjustment, it establishes the relationalmodel of micro virtual force and rotational angle; besides, according to the coverage rate ofoverall network, it can macroscopically control the adjustment of node angle; so it can adjustthe node sensing direction more reasonably. In the aspect of coverage redundancy control innetworks, it proposes the coverage strategy based on learning automata: according tooverlapping sensing rate and energy level of nodes, adopting LA for information exchangewith the surrounding environment, it can learn and choose the optimal nodes schedulingscheme of network coverage redundancy. Simulation results show that it can significantlyenhance the network coverage, and effectively control the network coverage redundancy.