Research

Iterative learning Control

Iterative learning control first appeared in 1980s. It is a mechanism that a system repeats the same task for many times and improves itself by learning from the past iterations. Iterative learning control can converge the error of a signal to zero after enough historical attempts. While for the traditional feedback control method, the error cannot be zero at all times. Iterative learning control method can be applied to the high performance mass production line such as laser scenery, 3D printer, and medical robotic arm.

Artificial Intelligence for Future Traffic Control

Traffic control is an essential aspect of the smart city scheme. The traffic data is collected and analysed via an intelligent decision engine to send out the real-time traffic signals. It can response to various incidents within the urban traffic network, and significantly increases the efficiency of the whole traffic network, such as achieving bus priority, minimum stop delays and average queue length reduction.

Neural Network based Decision Making

To make the appropriate decision in practice, it is significant to get access to as much desired information as possible. However, some key parameters of the target decision making problem are generally unknown or even time-varying. The decision makers are unable to make the reliable decision without knowing the real-time values of these parameters. The state-of-art technique machine learning is used to estimate the values of these unknown parameters using the recorded historical data. Using these estimations, the optimal control techniques such as model predictive control (MPC) provide the decision makers with optimal decision to achieve specific practical benefits, such as maximum net profit in industrial production.