Research On Single-step Delay Out-of-order Measurement Data Fusion Algorithm In UWSN

Underwater wireless sensor network (UWSN) is composed of a variable number sensor nodes with low-power and communication capability. Underwater sensor nodes are deployed in water randomly or regularly, and then they use their self-organization capability to form a network. The UWSN provides a new method for underwater environmental monitoring, but it uses acoustic communications which are different with traditional wireless sensor network (WSN). The UWSN brings many new challenges, such as positioning technology, network topology, routing protocols, acoustic communications and data fusion. The UWSN data fusion aims to reduce the amount of data transmitted, lower power consumption due to data transmission. It’s very important to extend the life cycle of the UWSN.Underwater acoustic channel is characterized by high propagation delay with dynamic changes, limited communication bandwidth, severe multipath effect and so on. Subjected to the constraints of its own physical conditions, there are usually different time delays in transmitting the measurement from the same target arrive out of sequence. Thus, "Out-Of-Sequence" Measurement (OOSM) phenomenon occurs. In the UWSN, how to handle the OOSM measurement directly related to the credibility of fusion results and prompts to improve data fusion theory.The OOSM data fusion problem is researched from single sensor and multiple sensors in UWSN, the main work is as follows:(1) Underwater acoustic channel is simulated with OPNET and underwater acoustic communications network model is established to analyze the network delay and observe OOSM phenomenon.(2) For single sensor OOSM problem, a one-step OOSM data fusion algorithm based on "backward prediction" is proposed. The algorithm solves single-step OOSM problem when measurement noise is correlated with process noise the same time. The process noise uses direct discrete-time model. The algorithm guarantees real-time, and filtering accuracy is higher than A1 algorithm’s. The simulation result shows its effectiveness. (3) For multi-sensor one-step OOSM problem, an OOSM data fusion algorithm based on equivalent measurement is proposed. This algorithm uses distributed manner. Equivalent measurement of one-step prediction is used to convert the OOSM data fusion to in-sequence measurement(ISM) data fusion. The partial one-step estimation is obtained by noise correlated Kalman filter, and then the global estimate is acquired by matrix weighted algorithm. The filtering accuracy of the algorithm is lower than direct update method’s, but the amount of calculation is reduced.