Algorithm On Task Allocation Of Sensor Nodes & Multi-Target Tracking In Wireless Sensor Networks

Target tracking is one of the important applications of Wireless Sensor Networks(WSN), and achieving efficient reliable and accurate mult-target tracking(MTT) has an important application value for the national defense and military, environmental monitoring, security surveillance, intelligent transportation and so on. Aiming at improving overall performance of MTT system, the research includes WSN-MTT nodes task allocation algorithm and WSN-MTT theory and algorithm,which has an important academic value and practical significance for the promotion of development and application of the networked measurement and control technology and strengthening cross of control science, computer science, communication and instrumentation subjects. The research work is funded by Guangdong Province Natural Science Foundation project (No.9151052101000013).Beginning with the basic steps of multi-target track in WSN, the compositive influence relationship between various tracking steps and tracking performance indexes is analyzed in the paper. The research progress at home and abroad of WSN-MTT nodes task allocation algorithm, WSN-MTT nets acquisition data fusion algorithms and WSN-MTT target state prediction algorithms to decide the research goals of this paper. Main research work in this paper are as follows:(1) WSN-MTT nodes task allocation formulation problem has been studied, FCM-MEMSOM and FCM-DPSO node task allocation methods have been presented, it pointed out the MTT node task allocation problem is a kind of cluster and many dimensions combinatorial optimization problem substantially. Aimed at random emerging targets in WSN-MTT monitor region and WSN-MTT nodes task allocation mathematical model to solve complexity. First achieving the unknown target number based on cluster separation distance threshold FCM, then guides particle to evlove thought based on the elastic neurons submodule receptive field dynamic adjustment mechanism and put tracking objective function as DPSO fitness function and lock quickly the optimal monitoring alliance, effectively reconciles the competition conflict in the task allocation of sensor nodes during the multi-target tracking, avoids further increase in energy consumption and prevents any possible real-time performance degradation due to the competition conflict, ensures tracking accuracy and reduces energy consumption. the simulation results show that FCM-MEMSOM method is superior to both the MEM neural network method and the nearest neighbor method, respectively with an energy consumption decrease by 7.4 ~14.5% and 8.2 ~ 27.9%, and with a task allocation time average reduces by 15.2% compare with MEM in the conditions of random uniform node topology. And also FCM-DPSO method is with an energy consumption decrease by 7.03% compare with MEM. Two methods have more advantages for WSN nodes deployment actual situation.(2) WSN data fusion algorithm based on kernel density estimation and nonparametric belief propagation is proposed. Full consideration of the unavoidable factors that random noise interference, sensor node perception vulnerability, network transmission, uncertainty actual application environment in WSN-MTT system, firstly, WSN-MTT system sampling data were accurately characterized by KDE for it can approximate any form density distribution function only by the sampling data; secondly, the acquisition data were precisely integrated without the sudden interference by gaussian mixture and Gibbs sampling fusion through collecting monitoring alliance data by NBP for its ability of dealing with distributed information. The KDE-NBP method is more suitable for many uncertainties of WSN, with a data fusion accuracy increases 34.1% compare with FCM method.(3) The WSN-MTT algorithm based on probability graph model and regularization particle filter (PGM-RPF) is proposed. Target motion process uncertainty factor is solved by PGM. Firstly,WSN-MTT target motion is considered as markov process, target tracking MRF models and target state probability distribution function are constructed, target state prediction problem is translated into MRF models implicit node probability deduction problems. Secondly, target state data from alliance node and other related node were fused accurately by NBP for its ability of dealing with distributed information. Thirdly, aiming at the problem of PF weights degenerate and loss of diversity particles, using RPF resample from continuous distribution instead of discrete distribution by putting the posterior distribution from discrete form convertting into successive density distribution form to improve the ablility about tracking some lower weights state. At last, target state accurate prediction and multi-target accurate tracking is achieved. The PGM-RPF method is respectively with tracking accuracy increases 54% and 3.5% compare with FCM-EKF and FCM-PF methods,(4) Synthesized WSN-MTT node task allocation, data fusion, state prediction, target tracking and so on. In this case, the performance of node task allocation and multi-target tracking method could be evaluated more systematically and expect to achieve good overall tracking results. Experiments proved that the feasibility and effectiveness of nodes task allocation and multi-target tracking methods in the practical environment application based on the developed WSN-MTT comprehensive experimental platform; Combining petrochemical enterprise, A petrochemical sulfur recovery installation employees tracking control system based on WSN is preliminary designed. Experimental results verify the effectiveness of the proposed algorithm. It can be extended to the complex environment and special applications monitoring tracking if the node properties, network coordination, remote access etc were improved.