Multistructure Radial Basis Function Neural-Networks-Based Extended Model Predictive Control: Application to Clutch Control

Abstract

In this article, the multistructure radial basis function neural networks (MSRBFNNs)-based extended model predictive control is further researched and discussed based on the work published in the previous conference. Taking the clutch control problem as an example, we first introduce the structure and training algorithms of the MSRBFNN in detail, and then, give a thorough derivation of the multidimensional recursive least-mean-square algorithm employed in the training process. Then, the controller design and stability analysis of the closed loop system are discussed in detail. Finally, the clutch control system model is refined, together with the simulation results and the experimental results on the test bench as verification. The results show that the proposed method is indeed effective in clutch control. As a general scheme is also proposed, it can be easily applied to other similar intelligent mechatronic systems, especially those with repeated tasks.

Citation

Recommended citation: C. Huang, X. Wang, L. Li and X. Chen, “Multistructure Radial Basis Function Neural-Networks-Based Extended Model Predictive Control: Application to Clutch Control,” in IEEE/ASME Transactions on Mechatronics, vol. 24, no. 6, pp. 2519-2530, Dec. 2019. (Paperurl)