Multi-sensor fusion based wheeled robot research on indoor positioning method

Abstract

In response to the poor positioning performance and errors of the wheeled robot under a single sensor, a combination of wheel odometry dead reckoning, Inertial Measurement Unit (IMU) heading angle information, and map environment information obtained from LiDAR is used for indoor positioning of the wheeled robot. By using data from multiple sensors, errors are reduced and positioning accuracy is improved. The results show that combining the Extended Kalman Filter (EKF), Gmapping algorithm, and Adaptive Monte Carlo Localization (AMCL) reduces the relative error by 8.35% compared to the single wheel odometry heading calculation method. Compared with the EKF fusion method for dead reckoning and inertial measurement unit heading angle information, the relative error is reduced by 4%, and there is no cumulative trend in angle error. This effectively addresses the challenge of cumulative errors in indoor positioning within environments that lack base stations, thereby enhancing accuracy and reliability in such specialized conditions.