RVNN

Name:

Control methods with neural networks
Neuro-Control

Type of course:

Lecture

Course content:

• Motivation and introduction

• Neural architectures and learning methods:

  Introduction, Multi Layer Perceptrons(MLP), simple perceptrons, explicit solution, gradient method, multilayer networks, generalization capabilities, the overfitting problem, backpropagation, momentum method, adaptive learning rate, Newton method, the Levenberg-Marquardt method, Dynamic Neural Networks, Time Delay Neural Networks TDNN, Recurrent Neural Networks RNN, Diagonal Recurrent Networks DRNN

• System identification:

  Introduction, Model structures, Linear model structures, Nonlinear model structures, Neural model structures, Experiment preparation and execution, Training of neural models

• NN-based control:

  Direct inverse control, direct training, model-based (indirect) training, control with internal model, control with feedforward, optimal control, online linearization, predictive control, feedback linearization: example, generalization, nonlinear continuous-time systems, nonlinear discrete-time systems, stability of nonlinear systems

Target group:

• 1st and 2nd level students from 5th semester onwards

Scope:

• 2 SWS lecture, 2 SWS exercise, 6 CP

Dates:

• The lecture takes place in the winter semester.

Documents:

• Announcement in the lecture

Proof of performance:

• Written exam

Lecturer:

• Dr.-Ing. Mohamed Ayeb

Back to overview