Research On Distributed Slam Based On Improved Particle Filter

Autonomous navigation is the process of autonomous mobile robots sensing surrounding environments, building the map, tracking object and obstacle detection, it is also the key technology which guarantee for robots working normally in known or unknown environment. This thesis focuses the research on autonomous navigation of robots based on laser radar; Simultaneous localization and mapping(SLAM) is a kind of effective method for autonomous navigation, and this thesis do further study on the method. To deal with the problem of centralized SLAM, such as the large calculate quantity, poor fault tolerate and stability, the thesis focus on the research of distributed SLAM based on particle filter, and several improve algorithm was put forward for the autonomous navigation. The research contents of this paper including:Firstly, this paper introduced the basic principle of SLAM and particle filter(PF), and emphasis on the realize and key process of SLAM, then introduce the relative technology, principal process and general implementation of particle filter, analysis on the model and implementation method of centralized SLAM, analysis the algorithm of centralized SLAM and the deficiency of the algorithm, finally introduce the algorithm of distributed SLAM which improve the above algorithm.Secondly, this paper puts forward the algorithm combined by particle filter and Particle Swarm Optimization Algorithm(PSO), and then the algorithm applied in the distributed SLAM system. The major works are those: 1) An improved proposal distribution has been obtained by particle swarm optimization algorithm dynamically. 2) On the basis of the above algorithm, proposed Quantum Behaviors Particle Swarm Optimization Algorithm(QPSO) improved the proposal distribution for improving precision. Using practical measurement data by Sydney University, simulation results show that the accuracy and stability of the algorithm have been improved.Thirdly, this paper replaced the normal particle filter of each local filter by the new filter which combination with Artificial Fish Swarm Algorithm, using the foraging behavior and huddling behavior of artificial fish to optimize the distribution of particles, and then proposed the improve algorithm. The simulation results show the precision and robustness of the improve algorithm.Finally, according to the sample exhaustion of traditional resampling technology, this paper puts forward the combinatorial optimization resampling algorithm of distributed SLAM based on particle filter, regulated the particles distribution of sub-filter witch need resampling by artificial fish swarm algorithm and fused particles, and this algorithm can deal with the problem of sample exhaustion effectively. Using practical measurement data by Sydney University, the simulation results show the validity of the improve algorithm, and compared with the existing algorithms.