An Adaptive Neuro-Fuzzy Control Approach for Motion Control of a Robot Arm

This paper proposes an adaptive Neuro-Fuzzy control approach for controlling the link variables of a 4 degree-of-freedom Selective Compliant Assembly Robot Arm (SCARA) type robot arm/manipulator. In the real world environment, the mathematical models of many robots are often not accurate, due to the presence of continuous disturbances that effect their dynamic equations, in addition to errors in parameter knowledge. Consequently, method that rely less on precise mathematical models are often preferred. One such Adaptive Machine Learning Technique is proposed to be applied here, for motion control of the robot arm. The controller uses an inverse learning Adaptive Neuro-Fuzzy Inference System (ANFIS) model only to train itself from certain given robot trajectories. Ideally, these trajectories should be obtained by directly measuring the robot arm responses for given inputs to capture the actual dynamics in the presence of all uncertainties. However, for algorithm validation, trajectories generated through simulations based on mathematical models assumed to be reasonably accurate, can also be used for the training purpose. This approach is used for design and implementation of an ANFIS controller which is shown to act work satisfactorily. Further possible developments of this method are also outlined.