No abstract

Inspite of so much research, key technical problem, naming chattering of conventional, simple and robust SMC is still a challenge to the researchers and hence limits its practical application. However, newly developed soft computing based techniques can provide solution. In order to have advantages of conventional and heuristic soft computing based control techniques, in this paper various commonly used intelligent techniques, neural network, fuzzy logic and adaptive neuro fuzzy inference system (ANFIS) have been combined with sliding mode controller (SMC). For validation, proposed hybrid control schemes have been implemented for tracking a predefined trajectory by robotic manipulator, incorporating structured and unstructured uncertainties in the system. After reviewing numerous papers, all the commonly occurring uncertainties like continuous disturbance, uniform random white noise, static friction like coulomb friction and viscous friction, dynamic friction like Dhal friction and LuGre friction have been inserted in the system. Various performance indices like norm of tracking error, chattering in control input, norm of input torque, disturbance rejection, chattering rejection have been used. Comparative results show that with almost eliminated chattering the intelligent SMC controllers are found to be more efficient over simple SMC. It has also been observed from results that ANFIS based controller has the best tracking performance with the reduced burden on the system. No paper in the literature has found to have all these structured and unstructured uncertainties together for motion control of robotic manipulator.Keywords Non-linear system Á Sliding mode control (SMC) Á Heuristic intelligent controllers Á Neural network (NN) Á Fuzzy logic (FL) and adaptive neuro fuzzy inference system (ANFIS)

Abstract-In this paper, an Evolutionary Optimized Neural Network (EONN) based control scheme is proposed. This control scheme is based on the fact that optimizing values of a few parameters of neural network can enhance its control performance. Radial Biased Neural Network (RBNN) is chosen here and PSO, one of the most emerging global optimizing techniques, is used to optimize the parameters of a RBNN. From hidden to output layer RBNN uses Gaussian function for mapping. Spread factor (s) of this intelligent RBNN is then optimized by a modified P SO to improvise its performance. The proposed controller has been verified by implementing it for position control of a robotic manipulator. For comparison purpose, proposed scheme has been verified with RBNN and the classical PD controller. MATLAB environment has been chosen for simulation study carried out. Robustness of the proposed controller has been checked by applying it to the manipulator for three different paths.

No abstract