Planning and Control for Autonomous Excavation

Abstract

This letter presents a novel planning and control approach for autonomous excavation that is independent of the soil composition and goes beyond a single dig. In contrast to the existing work, which is mostly based on position trajectories for the bucket motion, we define a single dig cycle directly by an end-effector force-torque trajectory. This trajectory applied to different types of soil, resulting in bucket motions that are used by expert operators to adapt to different terrain. Thereby, we can overcome the limitations of current approaches that suffer from the fact that soil interaction forces are dominant and immensely hard to predict or estimate. The end-effector forces and motion of the redundant excavator arm are controlled using a hierarchical optimization approach. Additionally, a large-scale iterative planner is proposed to consecutively execute single digs until a desired ground geometry is achieved. The shown simulation results include multidig excavations that precisely recreate the desired ground shapes. Practical tests are to follow.