A PSO based decentralized neuro-fuzzy-sliding mode controller for the twin rotor MIMO system

Abstract EN

In this paper, a decentralized neuro-fuzzy-sliding mode controller (NFSMC) with particle swarm optimization algorithm (PSO) is proposed to position the beam of the Twin Rotor MIMO System (TRMS) at the desired positions quickly and accurately.

Moreover, a synergistic combination of neural networks (NNs) with sliding mode control (SMC) methodology is presented.

In the present approach, in each sub-NFSMC (the first sub-controller is used to control the vertical sub-TRMS and the second used for the horizontal sub-TRMS) a feed-forward NN is used to compute the equivalent controller.

In addition, a Fuzzy logic controller (FLC) with a Mamdani fuzzy inference system is used as an approximation mechanism to compute the corrective term in the SMC.

The weights of each sub-net are updated such that the corrective control term of each sub-NFSMC goes to zero.

Moreover, the PSO algorithm is adopted to tune some parameters of each sub-NFSMC which are : the learning rate parameter of the back propagation algorithm (BPA) used to establish the multilayer feed-forward networks, the slopes of the switching surfaces used as inputs of each sub-FLC and output NN gains.

Finally, simulation results demonstrate that the controller designed can greatly alleviate the chattering effect and provide a very good accuracy and a very good response time, it also indicates that better performance can be achieved with this controller compared to the conventional SMC in the aspect of accuracy and control signals energy.