Study On Performance Optimization Of Monitoring In Wireless Sensor Networks

Wireless sensor networks bridge the gap between the physical world and in-formation world,enhance the ability to perceive the world,and open up a wide application prospect for the field of environmental monitoring,military reconnais-sance,medical care,intelligence house and so on.As a novel monitoring technique,the monitoring performance of sensor networks is a great concern in many appli-cations.In this paper,we consider about the issues of the sensing granularity,the sensing scope,the accuracy of node localization and the timeliness of event detection,and conduct in-depth study on the monitoring performance optimization of sensor networks in terms of theory,algorithm and implementation.The contribution is summarized as follows:(1)We have proposed a novel view-coverage model in video sensor networks which measures the coverage quality for the purpose of face recognition.Compared with traditional sensing model,view-coverage model considers about not only the position but also the orientation of the target,and therefore measures the cover-age performance with a finer granularity.Based on the view-coverage model,we study the problem of adjusting the orientation of individual nodes to improve the view-coverage ratio of the monitored area.We propose a distributed view-coverage enhancing algorithm(VCE),which improves the view-coverage ratio in a gradual manner by iteratively adjusting the orientation of individual nodes to increase their contribution degree of the view-coverage ratio.Simulation results show that VCE greatly improves the view-coverage ratio of the target area compared with random deployment and current adaptive algorithms.We also evaluate the coverage per-formance of the proposed algorithm by real deployed camera sensors in conjunction with face recognition system.(2)We have proposed a range-free virtual force-based localization algorithm(VFA)which achieves high accuracy in irregular deployed sensor network.We an-alyze the distance constraint between nodes by the connectivity information,and convert the localization problem into the problem of minimizing the inconsistency to the distance constraint.Inspired by the ideas of virtual potential field which is firstly proposed to solve the robot navigation problem,we establish the virtual force model according to the inconsistency to the distance constraint.During the process of the algorithm,we iteratively adjust the estimated locations of individual nodes-and gradually eliminate the inconsistency to the distance constraint.As a result,the localization accuracy is improved.The simulation results show that VFA al-most eliminates all the inconsistency to the distance constraint,and achieves higher localization accuracy in both regular and irregular deployed sensor networks com-pared with the state-of-the-art algorithms.Furthermore,we propose a probabilistic Q-UDG model to obtain the distance constraint between nodes in real deployed sensor networks,and apply it in VFA.The result shows that when the probability parameter takes a moderate value,VFA achieves a good result.(3)We have proposed a reliable and energy-efficient wake-up scheme to capture short-term events in structural health monitoring applications.In this scheme,a subset of nodes are equipped with vibration wake-up module,and act as sentry nodes.When the event happens,the sentry nodes are woken up by vibrations and they will in turn wake up the remaining deployed sensor nodes by broadcasting their radio signals.By analyzing the error rate of event detection and wake-up delay,we propose the sentry selection problem,and convert it into the probabilistic k-center problem,which is proved to be NP-complete.To this end,we propose a probabilistic greedy algorithm p-Greedy to solve the sentry selection problem.Simulation results show that p-Greedy achieves low wake-up delay with the reliability constraint of event detection.