Autonomous Robotic Mining Excavation Using Fuzzy Logic and Neural Networks

Abstract

This article describes our initial investigation into using neural networks and fuzzy behaviors for autonomous excavation control of a robotic front-end-loader-type mining machine. To utilize the experience and expertise from skilled human operators, a behavior-control approach based on fuzzy logic is developed. Nine typical behavior programs for general excavation tasks are constructed and implemented with fuzzy logic rules. Two neural networks are built to assess excavation situations and then select the corresponding behavior programs based on force/torque feedback data. Simple strategies for self-evaluation and fusion of fuzzy behaviors are presented. To verify the proposed approach, laboratory experiments are conducted using a PUMA 560 robot arm, a Zebra force/torque sensor, and a SUN workstation. Experimental results indicate that fuzzy behaviors are capable of reacting to unpredicted events during the excavation process and can complete the desired excavation goals successfully.