Fast Path Planning for Precision Weeding

Agricultural robots have the potential to reduce herbicide use in agriculture and horticulture through autonomous precision weeding. One of the main challenges is how to efficiently plan paths for a robot arm such that many individual weeds can be processed quickly. This paper considers an abstract weeding task among obstacles and proposes an efficient online path planning algorithm for an industrial manipulator mounted to a mobile robot chassis. The algorithm is based on a multi-query approach, inspired by industrial bin-picking, where a database of high-quality paths is computed offline and paths are then selected and adapted online. We present a preliminary implementation using a 6-DOF arm and report results from simulation experiments designed to evaluate system performance with varying database and obstacle sizes. We also validate the approach using a Universal Robots UR5 manipulator and ROS interface.