A Simultaneous Localization and Mapping Algorithm for Field Robots Based on Laser Range Sensor and Stereo Vision

Abstract

An improved simultaneous localization and mapping (SLAM) algorithm for a field robot is proposed in this paper. The algorithm enables the field robot to efficiently build a 2D and 3D graphic model of its environment based on the data acquired by a laser range sensor and a stereo vision camera module. Since reliable localization is an essential component of any autonomous vehicle, this research focused on the development of the methodology and experimentation of the SLAM algorithm that may be utilized in either indoor or outdoor environments. The sensor fusion design aspect and robot localization method are also presented in relation to the desired accuracy and possible agricultural applications such as greenhouse management and autonomous navigation in an open field. We emphasized in the derivation of the kinematic model of the field robot, particle filter, and path planning to enhance the performance of the SLAM algorithm and autonomous navigation. The experimental results have shown that the developed SLAM algorithm can deliver good accuracy in accordance with the initial uncertainty of the field robot. Based on the reconstructed map and the localization method, the developed field robot was able to navigate in an agricultural environment with desirable accuracy. Analyses of absolute and relative errors of field robot localization are discussed. It was also demonstrated that the developed SLAM algorithm successfully constructed and maintained maps for long runs.