#Custom harvester task planner manual

This paper introduces a fast and robust collision-free path-planning method based on deep reinforcement learning. In unstructured orchard environments, picking a target fruit without colliding with neighboring branches is a significant challenge for guava-harvesting robots. This motion and task planning framework has been successfully evaluated in simulation, and the real-time tests show that the harvesting task is accomplished with assured communication between sensing, planning and executing. The motion planning provides the abilities to the manipulator: to avoid the obstacles, to reach the targets, and to perform the detaching movement elaborated from a limited number of pre- defined strategies. The hierarchical task planning assures that the manipulator performs the harvesting task in the higher control level in corporation with other components: sensors system, mobile platform while dealing with the complicated and uncertain environment. The goal of this study is to introduce a framework for motion and hierarchical task planning which allows the manipulator to pick apples in the orchard.

#Custom harvester task planner manual

Selective harvesting for seasonal fruits like apples requires intensive manual labor in a short period however, due to the delicate property of the fruits, it cannot be performed with an existing commercial machine, which urged to be replaced by robots.