NXT* SCARA Model Based Design Controlled by Neural Network

Abstract

This paper describes experimental results applying Artificial Neu ral Networks (ANNs) to perform the position control of a real SCARA man ipulator robot. This approach has performed very successfully, with better results obtained with the Radial Basis Function (RBF) networks when co mpared to P controller and sliding mode positional controller. For mu lti -input mu lti-output (MIMO) continuous-time nonlinear systems, there are a few results available due to the difficulty in handling the coupling mat rix between different inputs. A stable neural network controller was developed for a class of nonlinear mult i-variable systems. The nonlinearit ies unknowns in the systems or in the controllers are approximated by linearly or nonlinearly parameterized neural networks, such as Radial Basis Function Neural Networks (RBF NNs) and Multilayer Neural Net works (M NNs). The introduction is talking about ANN wh ich can be learned by many methods to control non linear system SCARA and determine the effective co mputational technique structure simple as possible give high efficiency co mpared with classical P-controller. The NXT SCA RA modeling with two link planar robot arm, its kinemat ics and inverse kinematics with link motion equations, exp lanation for trajectory making of the edge of the robot arm.Then describes the controller design, its inputs, outputs and how the system tracking NN controller using RBFNN that is attractive for many problems, wh ich give rapid settling t ime, no overshoot and reduce error. The simu lation and results are exp lained which present the proposed NN imp roves the response and realizes good dynamic tracking and roboustness of system nonlinearity with self tuning NN without change the system parameters. The results are calculated by MATLAB/ SIMULINK.