An entry-exit path planner for an autonomous tractor in a paddy field

•A novel path planner for the entry and exit operations of a fully autonomous tractor was developed.•An occupancy grid map with virtual obstacles was designed and A* algorithm was applied on it.•Two path smoothing processes were used to reduce redundant turning and control waypoints.•Application of the entry-exit path planner to the paddy field was satisfactory. A field boundary with a single entrance as well as the entry and exit paths of the tractor based on the entrance location are key design requirements that must be considered when planning an effective path for an autonomous tractor operating in a paddy field to avoid collisions with the embankments that hold floodwater. In creating a complete in-field coverage path, an entry-exit path planner that generates paths to/from the start/end points of the auto-traveled path is needed to achieve effective field operations for a fully autonomous tractor. In this study, a novel path planner for the entry and exit operations of an autonomous tractor was developed using the A* algorithm to enable the tractor located at the entrance to automatically go to the start point of an agricultural task and return to the entrance after completing the agricultural task. An occupancy grid map with virtual obstacles was designed and the A* algorithm was applied to it to create a path for the tractor to reach its destination via the entry-exit path accurately in terms of positioning and heading. Two path-smoothing processes, a line-of-sight path smoother (LOPS) and a collinear-node path smoother (CNS), were performed to reduce redundant turning and control waypoints derived by the A* algorithm. The feasibility of using the developed algorithms was investigated via computer simulations followed by field tests with a 60-kW autonomous tractor. The simulation results confirm that the entry and exit paths generated by the proposed planner effectively guided the tractor to reach the given target points in terms of location and direction under the various entry, start, and end conditions by reducing the lateral and heading errors compared with those obtained without using the proposed search space and smoothers. For validation testing in two real paddy fields, the entry-exit paths were successfully generated in response to different entrance locations and the tractor followed the entry path and started the operation at the start point of the work path with a lateral deviation of ≤8.5 cm. In addition, the tractor navigated the exit