Controller Design Based on Wavelet Neural Adaptive Proportional Plus Conventional Integral-Derivative for Bilateral Teleoperation Systems with Time-Varying Parameters
Soheil Ganjefar,Mohammad Afshar,Mohammad Hadi Sarajchi,Zhufeng Shao
DOI: https://doi.org/10.1007/s12555-017-0739-1
2018-01-01
Abstract:In this study, a new controller method based on wavelet neural adaptive proportional plus conventional integral-derivative (WNAP+ID) controller through adaptive learning rates (ALRs) for the Internet-based bilateral teleoperation system is developed. The PID controller design suffers from dealing with a plant with an intricate dynamic model. To make an adaptive essence for PID controller, this study uses a trained offline self-recurrent wavelet neural network as a processing unit (SRWNN-PU) in parallel with conventional PID controller. The SRWNN-PU parameters are updated online using an SRWNN-identifier (SRWNNI) in order to reduce the controller error in realtime function. Using feedback linearization method and a PID controller, the presented control method reduced the tracking error in the subsystems of the teleoperation system, i.e., master and slave which are stabilized, respectively. Additionally, time-varying delay in teleoperation systems is considered as noise making the master signals be modulated because wavelt neural networks have a high susceptibility to remove the noise, thus the WNAP+ID controller is able to eliminate the noise effect. In this paper, we concentrated on the efficiency and stability of the teleoperation system with time-varying parameters through simulation outcomes. Moreover, the results of the WNNs are compared with those of multi-layer perceptron neural networks (MLPNNs).