A Control Architecture for Robotic Excavation in Construction

Abstract

This article presents a hybrid control architecture developed for robotic excavation. The lower-level controllers are designed using a combination of sliding mode control and fuzzy logic control. Control strategies at the higher level involve task decomposition in association with statecharts, and task execution and verification. Typical machine tasks are decomposed into subtasks and/or states. Graphical notation is introduced to facilitate software implementation of the control flow within statecharts. For each subtask, a statechart via its states at different priority levels selects to activate a number of corresponding lower-level controllers as a result from resolution of the machine resources and the control and data flows. The control architecture is designed with a view to managing hierarchical complexity, and facilitating the application of formal verification methods, software reuse, and lower-level control results. Field test results through the task of digging a trench are provided to demonstrate the feasibility of robotic excavation in moving toward construction automation.

High-level control methods for autonomous systems

“…Efforts need to be made to increase the level of automation of this important sector and to coordinate more of the involved processes in order to improve its productivity. For example, a hybrid control architecture for robotic excavation presented in [13] is one of such efforts. The lower-level controllers were designed by using a combination of sliding mode control and fuzzy logic control.…”

Earthmoving Construction Automation with Military Applications: Past, Present and Future

“…In [42], Matsuike et al (1996) developed an excavation control system, as a supporting system for large-depth excavation, in which the excavator was exactly positioned with the error less than 30-50 mm. A control architecture was developed in [43] for autonomous execution of some typical excavation tasks in construction. Using the same platform, Maeda [44] dealt with disturbances arisen in material removal process by proposing the Iterative Learning Control (ILC) with a PD-type learning function as a predictive controller to achieve a desired cut profile with nonmonotonic transients and converged faster by learning disturbances directly from command discrepancies.…”

Agent-based human robot interaction in a supervised autonomous system

“…The knowledge for it does not derive only from an expert operator; it can also come from the designer of the system. Fuzzy logic gives the most efficient rule-based representation that deals with continuous variables and fuzzy control at the higher layers of the architecture hierarchy can naturally enable the human machine interaction (11) and so be used as an addition to conventional control mostly implemented at the lower layers of the hierarchy (10) . For example, humans drive a car with no measurements and no computation receaving decisions at low control bandwidth; while high control bandwidth conventionally controlled units of the car system autonomously perform activites.…”