Localization And Objects Tracking Based On Particle Filter

These years the SLAM problem has been a hot issue in research on intelligent robot. It is a basic and important function to use the information captured from sensors to locate the robot reliably. It is also a challenging topic of much concern in the research of robot. Simultaneous Localization and Mapping (SLAM) is an essential capability for mobile robots to move in unknown environments. Detection and Tracking of Moving Objects (DATMO) is another key technology in the automatic control of robot. Early research divided SLAM and DATMO into two separate topics. In fact, they are complementary to each other. In order to make the robot's autonomy come true, such as AUV (Autonomous underwater vehicle), car, ship and plane to achieve auto-control function, SLAM and DATMO algorithms will merge together at last.The thesis first reviews the development and application of SLAM and DATMO and points out the relationship between them. Based on these basic concepts, the emphases of this paper are introduced: first is the SLAM algorithm based on Particle Filter. This algorithm researches deeply on the advantages of the Particle Filter applying on SLAM. The simulation result shows PF-based SLAM has higher location accuracy. Besides, it performs well on resolving non-liner problems. Compared with the simulation result of EKF-based SLAM, PF-based SLAM are much more accurate and robust. In addition, the PF-based DATMO algorithm is proposed too, the simulation result confirms the algorithm's accuracy and reliability. The experimental data shows that the accuracy and convergence of location are increased greatly with this algorithm to deal with robot location and tracking of moving object in dynamic environment. The comparison on consistency with EKF algorithm shows the excellent capacity of this algorithm to resolve the dynamic non-linear problem.The principle of this thesis is to apply Particle Filter to SLAM and DATMO. This approach factors the full SLAM posterior exactly into a product of a robot path posterior and landmark (static, dynamic) posteriors conditioned on the robot path estimate. In addition to sampling over robot paths and landmarks, PF-based SLAM samples over potential data of associations. Sampling over association data enables the simulation accuracy of PF-based SLAM to be greatly increased. Compared with traditional EKF algorithm, PF-based SLAM could get higher location accuracy. Besides, the PF-based DATMO algorithm is much more robust.