Experimental autonomous excavation using fuzzy behavior control

This paper presents a behavior-based control architecture for an autonomous excavator. The excavation model proposed is based on a front-end-loader type machine that excavates material composed mainly of blasted rock. At the lowest level of the control architecture individual fuzzy-behaviors provide a reactive interface between the control system and machine/environment. Behaviors consist of multiple actions or action sequences that are feasible to complete the task specified for each behavior. Task planning at the highest level uses finite state machines to decompose excavation goals into tasks and then tasks into behaviors. Features of the proposed control approach are discussed and compared with other behavior-based control approaches. Experimental results demonstrate the abilities of the system to effectively complete practical rock excavation tasks.