Design of visual odometry system for mobile robot

Abstract

This paper deals with design of hybrid software and hardware sensor, which can be used for mobile robots visual odometry task. The paper is focused to sensing raw data only. This approach combines onboard hardware sensors - accelerometer, gyroscope, magnetometer and camera with software module, which is mainly based on computer vision. Output from vision system is relative change of position and rotation of mobile robot. The position change is calculated for two axes based on vision data and the relative rotation is based on data fusion of three onboard sensors.