Abstract

Many agricultural robotics tasks require an end effector to hold stationary above individual plants in the field for short periods. Examples include precision harvesting, imaging and spraying. This effector may be mounted on a mobile base such as a large tractor or small robot, driving in the field. We consider how to optimise control of the base and the end actuator together, to minimise total time taken to visit the plants. Our approach is based on low level combination of simple motion primitives, with mid level target clustering, and higher level planning. For the high level, three strategies are compared and evaluated in simulation: baseline stop-and-spray, constant velocity, and variable velocity. The baseline strategy is common in existing systems, and is shown to be outperformed by the new methods. The application considered here is weed spraying, but the methods are applicable to many tasks.