Gang Cao

Citation

Thesis Title
Gaussian Process based Model Predictive Control

Read article at Massey Research Online:

Established predictive control theories and methods are primarily designed for dynamical systems that are deterministically described by mathematical or data-driven models, rather than a probability distribution. With advanced control and machine learning theories, Mr Cao developed new predictive control algorithms with a guaranteed stability to handle control issues where unknown dynamics are described by probabilistic Gaussian process models. These algorithms allow evaluating the model uncertainty by obtained variances and directly taking into account the control problem. In addition, a linearisation of Gaussian process models was proposed to solve the control problem efficiently. Mr Cao's research has provided a valuable insight into the probabilistic control theory based on advanced control and machine learning.

Supervisors
Associate Professor Fakhrul Alam
Associate Professor Edmund Lai