Decentralized Fusion with Relative Measurements for Delayed Sensor Networks

Multisensor data fusion based on wireless sensor networks (WSN) has many advantages over the traditional data fusion and is faced with many challenges. Because sensor nodes in WSN decentralize over wide area and data must be transmitted to the central processor by use of special wireless transport protocols for fusion, data delay and missing continually occur across sensor networks. Therefore, sensor measurements arrive in the central processor out-of-sequence, accordingly, optimal fusion problem for "Out-of-sequence" measurements (OOSM) is presented. For OOSM fusion problem, two main methods are presented such as giving up directly and reused delayed information. Some problems will be produced such as lower fusion accuracy and bad convergence and stability for the former. For the idea of reusing delayed information, some effective fusion methods have been reported, whereas many limitations still exist in them, for example confined application, huge computation and bad implementation and so forth. In order to overcome above problems, a novel decentralized fusion estimator based on relative measurements, which is to reuse the delayed OOSM, is proposed in this paper. Through making use of basic systemic state equation, the relative measurements of current state corresponding to delayed measurements can be obtained and fusion can be performed by use of federal fusion. In addition, some other forms of the decentralized fusion estimator are all given and discussed for different situations. The algorithm analysis shows that the decentralized fusion estimator is suboptimal and its fusion accuracy is lower than that of current batch or centralized fusion, but it has better practicality and lower computation. Finally, simulation experiment proves the validity of the proposed decentralized fusion estimator.