Automated guided vehicle robot localization with sensor fusion

Abstract

Robot localization is vital for the operation of an automated guided vehicle (AGV) but is susceptible to problems such as wheel slip. With more sensors fused together, the more environmental information can be collected by the AGV, which helps with the localization of AGV. Inertial measurement unit (IMU) and global positioning unit (GPS) are usually implemented to improve robot localization but are susceptible to noise and are effective outdoors. Indoors, however, are more suitable with light detection and ranging (lidar) device. This paper implements extended Kalman filter (EKF) and unscented Kalman filter (UKF) for robot localization on AGV. AGV localization was tested with EKF and UKF on three different test tracks with different turn conditions. The performance of the EKF and UKF was compared to each other. Different sensors were implemented along with sensor fusion. UKF generates better odometry estimation than EKF with 24.07% better accuracy. With the usage of lidar, wheel slip was compensated.