Method for improving the accuracy of mobile robot odometry based on machine learning

Abstract

Omnidirectional mobile robots have the characteristic of moving in any direction with three degrees of freedom in plane XYZ. The positioning accuracy of a mobile robot is not only affected by its manufacturing errors but also may be affected by vibration, load, and other factors, resulting in errors in the process of moving. To reduce the influence of the above factors, this paper presents an error compensation method based on Gaussian process regression (GPR) for omnidirectional mobile robots. The method requires fewer learning parameters and training points, which enables more computational efficiency in the learning process. To verify the effectiveness of the method, a simulation experiment is carried out. The simulation results show that the accuracy of the odometry in the three degrees of freedom of the plane is improved by 89.70%, 88.09%, and 96.32% respectively.