Vehicle Position Tracking Using Kalman Filter And CWLS Optimization

Vehicle position estimation for wireless network has been studied in many fields since it has the ability to provide a variety of services, such as detecting oncoming collisions and providing warning signals to alert the driver. The services provided are often based on collaboration among vehicles that are equipped with relatively simple motion sensors and GPS units. Awareness of its precise position is vital to every vehicle, so that it can provide accurate data to its peers. Currently, typical positioning techniques integrate GPS receiver data and measurements of the vehicles motion. However, when the vehicle passes through an environment that creates multipath effect, these techniques fail to produce high position accuracy that they attain in open environments. Unfortunately, vehicles often travel in environments that cause multipath effect, such as areas with high buildings, trees, or tunnels. The goal of this research is to minimize the multipath effect with respect to the position accuracy of vehicles. The proposed technique first detects whether there is disturbance in the vehicle position estimate that is caused by the multipath effect using hypothesis test. This technique integrates all information with the vehicle's own data and the Constrained Weighted Least Squares (CWLS) optimization approach with time difference of arrival (TDOA) technique and minimizes the position estimate error of the vehicle. Kalman filter is used for smoothing range data and mitigating the NLOS errors. The positioning problem is formulated in a state-space framework and the constraints on system states are considered explicitly. The proposed recursive positioning algorithm will be comparatively more robust to measurement errors because it updates the technique that feeds the position corrections back to the Kalman Filter as compared with a Kalman tracking algorithm that estimates the target track directly from the TDOA measurements. It compensates the GPS data and decreases random error influence to the position precision. The new techniques presented in this thesis decrease the error in the position estimate. Simulation results show that the proposed tracking algorithm can improve the accuracy significantly.