SLAM Based On Improved Particle Filter In Dynamic Environment

A key to realize the fully autonomous mobile robot is how to solve the problem of simulation localization and mapping (SLAM), which is the important foundation for the mobile robot to achieve environment exploration and navigation. With the development of sensor, how to describe the information from sensors effectively and how to choose the appropriate system state estimation scheme under a dynamic scene is the key problems need to be solved the localization and map building.Based on previous works, this dissertation studies on simultaneous localization and map building of mobile robot. For the map expression of indoor semi-structured characteristics and the traditional method of adaptability to dynamic environment, some improved algorithms are put forward to improve the accuracy of SLAM algorithm, to maintain the mapping consistency. The main work includes:1. A relative localization method is proposed, which combines the self-localization results using scan matching and the estimation results using wheel odometry. Due to the discrete characteristics of the measurement of the laser range sensor and the characteristics of local line segment showed by indoor observation, the point-to-line iterative closest point method is used in local matching to correct the error of wheel odometry. In order to adapt the real-time performance of the matching algorithm to SLAM, a hybrid fast search algorithm is adopted. Using the fast corresponding point strategy to accelerate the solution of the error function, so as to improve the accuracy of the motion model to improve the prediction accuracy of the SLAM algorithm.2. In view of the structural characteristics of the indoor semi structured environment, an optimal mapping and localization algorithm based on cellular automata (CA) is proposed. By using the space divided by occupied grid, the cellular automata is introduced into the traditional SLAM iteration cycle, and the CA-enhancement is carried out in the map updating process, which improves the consistency of mapping and thus optimizes localization results3. For the problem of map modeling of dynamic scene and the extension of the cellular automata SLAM algorithm, a hierarchical reinforcement mapping and localization algorithm is proposed to adapt to dynamic environment. The algorithm inherits the dynamic and static description of the grid map, and adds static and dynamic state values by statisting the observation frequency of the grid attributes to describe mobile property of an object effectively. In the meanwhile, Using space hierarchical difference to deal with abundant scan information, the common dynamic SLAM problem is transformed into the mapping problem of the structural environment and the construction problem of the mixed map of the dynamic environment. The SLAM algorithm integrated with CA rule is applied to the dynamic scene effectively.Based on the laser range finder as an environmental sensor, the algorithm is tested on Gazebo simulator and the youBot mobile robot using ROS. The experimental results show that the proposed method is effective and reliable.