Tantárgy adatlapja
The goal of this course is to highlight important results in nonlinear systems’ and control theory and robotics. The subject of this course will be discussed in weekly consultations on the following topics:
I. Model Predictive Controller (MPC) techniques [1], [2]
- Basic MPC techniques employing linear prediction models
- Receeding horizon control, offline control, and preliminary control feasibility test
- Nonlinear MPC (NMPC) and convex approximations
- Stochastic (N)MPC and convex approximations
- Unknown-input reconstruction, parameter estimation, and state observation in an MPC framework
II. Gaussian process (GP) regression models [3], [4]
- Multivariate function approximation from measurements using Gps
- Hyperparameter tuning using log-likelihood optimization
- Sparse GP approximations: FITC, VFE, distance-based dictionaries
- Dynamic model calibration using GPs
- Prediction models augmented with Gps
- Stochastic GP model-based predictions using different model approximations (e.g., Taylor approximation or analytic moment calculations)
- MPC approaches employing nonlinear prediction models augmented with GP regressors (GP-MPC)
- GP-MPC implementation techniques and approximations:
III. Useful numerical methods, available software tools, and their usage
- Advanced Matlab/Simulink tools and techniques relevant in the field
- Algorithmic differentiation tools (e.g., CasADi) and gradient-based nonlinear optimization solvers (e.g., IPOPT, FORCES PRO)
- GPML Toolbox for GP manipulation
