A number of researchers have investigated the feasibility of automating excavation. Many of these studies have addressed the possible use of autonomous excavators during unmanned phases of establishing manned Lunar or Martian research stations 3, 4. Much of the work on terrestrial excavation has focused on teleoperation, rather than on the system requirements for autonomous operation. Although there have been a number of valuable theoretical and experimental contributions to the field of autonomous, robotic or teleoperated excavation 5-8, autonomous operation of a full-scale excavator has not been commercially demonstrated.Many of the experimental studies reported in the literature involve using conventional industrial robots fitted with buckets to excavate in a bed of loose sand. While there are parallels between “classical” robotics and robotic excavation, there are also some pronounced differences. In particular, an excavator is not fixed relative to the work piece; it plastically deforms the work piece by applying large forces and is caused to move relative to the soil by the same large forces. Furthermore, strategic and bucket trajectory planning must necessarily occur in a dynamic environment; if the excavator is not changing the profile of the soil being worked, it is not doing useful work.This article presents some results of the autonomous excavation project conducted at the Australian Centre for Field Robotics (ACFR) with a focus on construction automation.The application of robotic technology and computer control is one key to construction industry automation. Excavation automation is a multidisciplinary task, encompassing a broad area of research and development• planning• monitoring• environment sensing and modeling• navigation• machine modeling and control.The ultimate goal of the ACFR excavation project is to demonstrate fully autonomous execution of excavation tasks in common construction, such as loading a truck or digging a trench. A number of difficult theoretical and practical problems must be solved to achieve this objective. The problems fall into three main groups: excavation planning, sensing and estimation, and control.