Row-following accuracy of an autonomous vision-guided agricultural vehicle

This paper seeks to establish the accuracy of an autonomous vehicle working in a field of transplanted cauliflowers. The main sensing systems, odometry and image analysis, are briefly described as is the control system which is based on a fusion of the two data sources using a Kalman filter. Experiments to establish accuracy are described. These were carried out on four plots of cauliflowers with varying degrees of disruption to the visual scene. The RMS error of vehicle lateral position control was 20 mm, while the RMS error of estimated vehicle position was about 10 mm. Little effect of the disruption on position control was observed, It is concluded that these accuracies would be sufficient to control a vehicle and an associated crop treatment device but that improvements to the vehicle controller would make the control of the treatment device easier.