EKF-Based Position Estimation of Wheeled Mobile Robot for 2-D Motion

Abstract

The fundamental requirement for a mobile robot is to determine the accurate information of location in any environment. It is well established that the estimation of the robot's true dynamic state using the system model alone is very difficult. In the present paper, the sensor measurement data along with the system model has been used to determine the true state of the system using sensor fusion technique using the extended Kalman filter, a two-step recursive filtering technique. The prediction step is implemented using the kinematic model of the robot while correction step directly uses the perception and minimizes the difference between relative and absolute positioning to correct the mobile robot’s pose estimation. In the present work, a 2-D environmental scene has been designed in CoppeliaSim simulator incorporating sensor information fusion model and robot motion model for mobile robot. The robot motion model is introduced by using the wheel odometry method and ultrasonic sensor for localization in a known environment. The extended Kalman filter is used as a multi-sensor data fusion technology to obtain optimal robot localization for 2-D motion. The results of system model and sensor fusion model are presented and compared. The sensor fusion model shows better estimation and localization of the robot.