Research On GPS/DR Integrated Navigation Algorithm Based On Particle Filter

As a key technology of Intelligent Transportation System, vehicle positioning is of vital importance to improve transportation efficiency and enhance transportation safety. Global Positioning System, named GPS, is most commonly used among many methods for vehicle positioning. Despite the advantages of simple operation, high precision and wide coverage, GPS is restricted in some conditions and needs to cooperate with other ways of positioning. Dead Reckoning, another easy method for vehicle positioning, can calculate current position with angular velocity, velocity and previous position of a vehicle. However, error caused by measuring devices in DR will accumulate over time. Combining GPS and DR will overcome shortcomings of each other and greatly improve positioning accuracy. GPS/DR integrated navigation is now commonly used as a positioning method.It’s crucial to choose an appropriate data fusion scheme in GPS/DR system. Common data fusion algorithms include Extended Kalman Filter, Unscented Kalman Filter, Federated Kalman Filter etc. Improved Particle Filter is proposed and applied to GPS/DR integrated navigation. Main work done is as follows:First of all, the development and status of integrated navigation and data fusion technique are introduced. Secondly, the basic theory of integrated navigation including error analysis and model establishment of GPS/DR system and the basic principle of KF, EKF, UKF, PF is studied.Thirdly, both shortcomings and improved methods of PF are studied. Aiming at the disadvantage of large amount of calculation, global sampling is applied to EKPF. To choose a better proposal distribution, Mixture Kalman Particle Filter combined with Strong Tracking Filter is proposed. Moreover, an improved PF combining the two aspects is proposed and studied.Then, with MATLAB tool, both the improved algorithms and the existing algorithms are simulated on GPS/DR model established above. Simulation results of these algorithms are compared in aspects such as estimation error of eastern position, eastern velocity, northern position and northern velocity and time consumption, with which it is clear that whether the improved algorithms proposed are effective.At last, main work done is summarized and future research is prospected.