On The Fault-Tolerant Target Detection Algorithms In Wireless Sensor Networks

With the fast development of sensor networks, many new applications emerged in recent years. In these applications, dynamic target tracking is one of the most important issues. Target tracking is extensively desirable for diverse field of subjects in natural science, wild animal study and military information collection, wherein the target detection is the precondition and basis for the target location and target tracking. So far there have been a number of research results in the literature on the target detection algorithms for wireless sensor networks. Nonetheless most of the existing algorithms address exclusively the accuracy of target detection, but ignore the robustness against the faulty nodes in networks. However, in real applications, Byzantine faults may happen due to the harsh environment, where the nodes are assumed to send inconsistent and arbitrary values to other nodes during information fusion. Obviously, fault-tolerant target detection algorithms are highly desired to achieve the robustness to the Byzantine faults. Obvisouly, the fault-tolerance will become a crucial problem in target detection for wireless sensor networks.In this paper, the development of sensor network and the challenges to be faced, the application fields the WSN together with the key technologies of sensor network are reviewed at first. Then some known algorithms for target location and tracking in wireless sensor network are discussed to highlight the crucial problems therein.As an important target detection algorithm, the Value Fusion Algorithm are addressed, wherein the measurements of all nodes are collected through OM algorithm and the n extremes are discarded to exclude the abnormal readings by Byzantine faulty nodes. The problem of the VFA scheme is the cost of communication overhead, which will become a critical problem when the number of nodes is large enough. Due to the strict energy constraints in wireless sensor network, communication efficient algorithm is desirable. An energy-efficient HBA algorithm was proposed recently, wherein SM algorithm instead of OM algorithm are utilized hierarchically. Compared with the OM algorithm in the whole interest region, hierarchical SM algorithm in HBA scheme is able to reduce the energy dissipation significantly. However, the hierarchical processing sometimes brings about a number of communication rounds, thus giving rise to an unwanted long time delay.In this paper, the hierarchical value fusion scheme (HVFA) is proposed based on both the VFA and HBA algorithms. Just like the HBA algorithm, all sensor nodes are partitioned into several subgroups and the SM algorithm is performed hierarchically among nodes to correct faults. However, unlike the HBA scheme where the SM algorithm will be performed thrice among the subgroup heads, the member nodes in each subgroup, and the shared nodes between subgroups when faulty head is detected, respectively. The SM algorithm will be performed only twice among subgroup heads and when shared nodes correct the faulty head in the HVFA scheme. As a result, the number of communication rounds in HVFA is much less than those in the HBA. It is validated through numerical anlaysis and computer simulations that, the HVFA algorithm can enable energy-efficient target detection with much less processing time delay compared with that of HBA algorithm.As for VFA algorithm, one of the challenges is the determination of the number of extreme values to be excluded due to the "mean value" calculation during measurement exchange among neighboring nodes. As a comparison, we explore the utilization of "median value" to effectively remove those extreme values for target detection. In median value fault-tolerant algorithm, the median value of all neighboring readings is used to approximate the local observation to the possible present target. And it is validated through the simulations on the OPNET platform that, the algorithm can reduce the energy consumed and achieve a good performance of fault-tolerance and detection probability when the number of Byzantine faulty nodes is small and the node density is large enough.