Autonomous navigation using a robot platform in a sugar beet field

Abstract

An RTK-DGPS (Real Time Kinematic Differential Global Positioning System) based autonomous field navigation system including automated headland turns was developed to provide a method for crop row mapping combining machine vision, and to evaluate the benefits of a behaviour based reactive layer in a hybrid deliberate systems architecture. Two experiments were performed at the same time: following of pre-planned paths reconstructed from crop row positions based on RTK-DGPS and crop row mapping by combining vision-based row detection with RTK-DGPS information. The standard deviation, mean, minimum and maximum lateral error of the robot vehicle while following a straight path on the field with RTK-DGPS at a speed of 0.3 m s−1 were respectively 1.6, 0.1, −4.5 and 3.4 cm. The standard deviation, mean, minimum and maximum of the heading error were 0.008, 0.000, −0.022 and 0.023 rad. The point-in-polygon algorithm proved to be a suitable method for detection in which part of the field the actuator position coordinates or the field of view of the camera are located. A smooth headland path that connected to the subsequent path along the crop was generated in realtime using a spline based algorithm. The hybrid deliberate software architecture with a behaviour based reactive layer allowed a convenient evaluation of the robot performance. Results from the field experiments showed that the implement can be guided along a defined path with cm precision using an autonomous robot navigating in a field.