Research On Multi-robots Cooperative Localization Algorithm Basd On GPU Accelerated

The research and development of mobile robot technology not only have broad market demand,but also have important social significance.Compared with a single robot,multi-mobile robot can achieve better work efficiency and more robust.The real-time and accurate positioning of each robot in the multi-mobile robot system is the basis of the whole system's safe and efficient work.Cooperative localization of multi-mobile robot makes full use of the measurement information between robots to further improve the accuracy and robustness of localization.To improve the real-time performance is one of the key points in the research of multi-mobile robot cooperative localization.At present,GPU acceleration has achieved good results in deep learning,image processing,scientific computing and other fields Therefore,this paper studies how to improve the speed of multi-mobile robot cooperative localization algorithm with GPU acceleration technology.The main results of the research work are as follows:(1)Aiming at the particle filter algorithm for single robot localization,combined with the characteristics of GPU hardware,the parallelism of the particle filter algorithm is analyzed,and a parallel particle filter positioning algorithm based on CUDA is proposed.The simulation results show that the parallel algorithm can achieve about 4.6 times acceleration effect when used 1024 particles,and with the increase of the number of particles,the acceleration effect of the parallel algorithm is more obvious.(2)In this paper,the cooperative localization algorithm of multi-mobile robots,which is based on Gibbs sampler based cooperative particle filter(GSCPF),is studied.A multi-mobile robot cooperative localization experiment platform based on UWB technology is constructed,and three Turtlebot2 robots are tested.The experimental results show that the positioning accuracy of GSCPF is about 40%higher than that of extended Kalman filter.(3)In order to improve the real-time performance of GSCPF algorithm,based on GPU hardware and CUDA Programming Model,the parallelism of GSCPF algorithm is analyzed,and a parallel GSCPF algorithm is proposed.The experimental results of three Turtlebot2 mobile robots show that the running speed of the proposed algorithm is about 4.8 times of the original GSCPF algorithm,and the simulation results of seven mobile robots show that the running speed of the proposed algorithm is about 9 times of the original GSCPF algorithm.It showed that the GPU acceleration algorithm has a good acceleration effect,and with the increase of mobile robots in the system,its acceleration effect is more obvious.