Method For Localization In Wireless Sensor Networks Based On Belief Propagation

Positioning technology is one of the critical technologies for wireless network applications in the commercial, civilian and military, and data in network without location information is often meaningless. Wireless Sensor Networks (WSN) has been the rapid development in recent years. WSN has features of low power, low cost, short time delay, reliable and security. Meanwhile, WSN has a wide range of needs for self-localization. However, Global Positioning System, the most widely used positioning system, cannot fully meet the requirements of WSN node calibration, since the weak indoor signal, low positioning accuracy and higher power consumption of GPS. Therefore, technology of self-localization in WSN came into being.In order to achieve high-precision location of sensor in WSN, this thesis conducted a in-depth research in self-localization method in WSN. Currently, self-localization method is divided into range-based positioning methods and range-free positioning methods due to whether the localization is based on the measurement. As the accuracy of range-based positioning methods is generally higher than range-free positioning methods, this thesis pay more attention to the former. Time Difference of Arrival (TDOA) is a high accuracy ranging technology without need of time synchronization, and it is widely used in fields of node localization in WSN. This thesis presents an EC-TDOA (Error-Checking TDOA) ranging method. EC-TDOA gets a more accurate distance measurement result by error correction application of experimental data and least-squares regression on TDOA. Graphical model of sensor localization based on EC-TDOA is build, and belief propagation (BP) algorithm and non-parametric belief propagation (NBP) algorithm is applied for achieving the maximum a posteriori probability and sensor location. In order to verify the actual effect of positioning methods, a positioning system of WSN is designed and implemented. Positioning system make nodes using chips meet the standard of ZigBee protocol. In addition, host computer software is complimented on.NET Framework, and performing real-time calibration of sensors in WSN. According to the positioning system data collection, MLE(maximum likelihood estimation) based on TDOA positioning method, NBP based on TDOA positioning method, MLE based on EC-TDOA positioning method and NBP based on EC-TDOA positioning method are applied to experiments. The results are discussed and analyzed from multiple indicators such as RMSE (Root Mean Square Error) and CDF (Cumulative Distribution Function). The results show that the NBP based on EC-TDOA positioning method can be achieved. The maximum positioning error, the average location error and the positioning frequency are 4.3cm,1.64cm and 17 times/sec. The positioning performance is better than the other three kind of positioning methods.Compared with the MLE positioning method, NBP based on EC-TDOA positioning methods decrease errors from distance measurement, and realize the positioning accuracy of quadratic optimization by using NBP. It finally improves the positioning accuracy of the results. The application of NBP based EC-TDOA positioning method shows features of real-time performance, robustness. Furthermore, it reduces the fluctuation of sensor measuring position significantly, and plays a filtering effect.