Research On SLAM And Navigation Algorithms Of Multi-sensor Fusion In Mix Scenes

With the rapid development of sensor technology,computer science and artificial intelligence technology,robotics is advancing at unprecedented speed and scale.In the process of robot development,intelligence has gradually become an important indicator for measuring robots.The robot’s simultaneous localization and mapping(SLAM),path planning,and ability to avoid obstacles are typical characteristics of robot intelligence.Based on this background,this subject mainly conducts research on SLAM and navigation algorithms based on multi-sensor fusion positioning.The main work is as follows:(1)Firstly,the model of the robot system is carried out,and the analysis mainly involves the following aspects: the motion model of the robot,the sensor model and the environment perception model.These models are the basis for the realization of robot positioning in unknown environments and the construction of environmental maps.(2)For the mobile robot positioning technology,according to the extended Kalman filter theory,a fusion positioning algorithm based on IMU and odometer is designed.Aiming at the simultaneous localization and map construction technology of mobile robots,based on the particle filter theory,a filter-based SLAM algorithm is studied.Starting from the idea of Bayesian filtering,the theory of particle filtering is deduced.According to the resampling theory,the importance function is introduced to make the sampling more concentrated in the peak area of the probability density function and reduce the influence of particle degradation.Aiming at the inefficiency of particle filter in high-dimensional state space,the concept of Rao-Blackwellized particle filter is introduced.On the basis of RBPF,the proposed distribution and selective resampling are improved,and the GMapping algorithm is introduced.(3)For the path planning problem of mobile robots,three global path planning algorithms,namely A*,Dijkstra and RRT algorithms are compared,and the theoretical derivation of these three algorithms is carried out.In the derivation of the A algorithm,by introducing the heuristic function,a more efficient path search is realized,and it is pointed out that the A* algorithm can quickly find the global optimal solution,and has high efficiency and precision.In the derivation of the local path planning DWA algorithm,by analyzing the motion characteristics and constraints of the robot at different speeds,the optimal solution of the velocity and angular velocity is deduced,and the path planning is combined with the robot dynamics model.Based on the global path,the DWA algorithm evaluation function is improved,which combines global path planning and local dynamic window search,so that the robot can quickly respond and adjust actions in unknown environments to achieve safe and efficient navigation.(4)In the real environment,the robots with high-performance,high-cost and lowperformance,low-cost are used as the robot experiment platform,and the mapping of the unknown environment is completed based on GMapping SLAM.On the basis of the constructed environment map,the path planning algorithm is experimented.The algorithm is transplanted to the low-cost and low-performance AGV to verify the reliability and robustness of the algorithm.