Development of an Agricultural Mobile Robot using a Geomagnetic Direction Sensor and Image Sensors

Abstract

To solve the problem of labour shortage and the decrease in skilled agricultural workers, it is thought that mobile robots, which substitute for conventional tractors, will play an important role in the next century. This paper describes a mobile robot system, including a positioning system, using image sensors. A mobile robot which can be controlled by positions obtained from the image sensors and the heading angles from a geomagnetic direction sensor (GDS) has been developed. Because of its low cost, the GDS has recently been used in much research related to mobile robots. However, there are two major problems in using the GDS as a sensor: the error from magnetism which exists around the robot and the error from the robot inclinations. Owing to these error factors of the GDS and in order to utilize it effectively, a method is developed in this paper to refine the GDS output using neural networks (NN). The positioning system for the mobile robot is based on the principle of triangulation; it is composed of a main system and two subsystems both having a static image sensor. Each subsystem is able to follow the movement of the robot and to measure the angle to the robot using image analysis. The main system calculates the robot position using two angles from the subsystems and sends it to the robot using a telecommunication system. Finally, assuming that the robot will transport hay on a meadow, a control algorithm for the mobile robot system is developed, and a field test is conducted on grassland to evaluate the developed robot system. The average error of the final position for each target position was found to be about 0·4m. The absolute maximum error and the r.m.s. error of the position for the predetermined path were about 0·51m and 0·23m, respectively, over all travels.