The Development of an Autonomous Vehicle for use in Agriculture

Abstract

The purpose of this research is the development a versatile autonomous research vehicle as well as a machine vision row guidance system for the research vehicle. The vehicle developed will be relatively small compared to conventional agricultural machines with the vision of more and smaller vehicles being used in the future. Smaller autonomous vehicles are able to be more versatile and site specific as well as having lower input costs, but are only feasible if they are autonomous, since the cost of labor offsets the costs of technology. A two wheel differential steering vehicle was selected and adapted for autonomous use by incorporating electronic actuators into the conventional steering linkage. The machine vision was developed using a small CCD camera that was programmed to scan images to find high contrast pixels and locate a line or row in an image. Error measurements from the machine vision system were reported to an onboard computer, which was programmed with a Proportional-Integral-Derivative controller to settle errors. The system used an 'aggregate' error approach, accounting for both angular and offset errors. The system was tested in an area of consistent illumination in various situations and compared with a human operator's performance. Test situations include straight and curved paths as well as discontinuities of the followed path. The system performed well in most situations with standard deviations similar to that of the human operator. The use of a single sensor for autonomous vehicle control was shown to be unfavorable and inadequate. Even so, the possibility of autonomous vehicles with multiple sensors is promising even though safety considerations remain to be solved before true autonomous operations can be introduced in agriculture.