Visual odometry for an outdoor mobile robot

Abstract

This paper presents an alternative solution to the problem of estimating the motion of an outdoor mobile robot through odometry. The solution consists of the use of vision and pseudo-optical flow techniques to make motion estimates of a camera attached to the robot chassis. The use of vision makes the system independent of locomotion (for example, the differences between a robot with wheels and one with legs) and is shown to be applicable to a variety of outdoor ground surfaces, such as tarmac, gravel, grass and dirt. Finally, the simplified calculations result in high processing speeds on moderate equipment, with experiments showing that high speed motion (approximately 300 degree per second rotations and 350 millimetre per second translations) can be reliably estimated.