The Research of Neural-Fuzzy Inference System Model for Arm of Excavator Robot

Abstract

To improve the trajectory planning control accuracy of the working device of excavator robot when working for excavation, reduce working device of excavator robot to be the two joints two dimension robot arm composed of arm and bucket when analyses. Creating inverse kinematics model, need to relate to terminal position and orientation space of bucket and joint space and cylinder space of working devices to plan trace, so that to control excavator robot in each space. Because of the geometry complexity of inverse kinematics, the nonlinear of electro-hydraulic system, the uncertainty and non-uniqueness for inverse mapping, lead to the difficulty of trajectory planning. To improve the precision for tracking desired trace, obtaining inverse kinematics mapping between terminal trace and joint angle, use two adaptive neural-fuzzy inference system(ANFIS) to learn inverse mapping relations between (x, y) two joint coordinate and joint angle, create ANFIS inverse mapping model. Select I/O curve data of inverse mapping to train ANFIS structure, obtain I/O mapping curve of fuzzy model, realize the aim of obtaining corresponding joint angle based on given desired excavation trace. Use the model to trace expected motion trace finally, indicate by simulation that trace precision can meet the actual demands.