Research

Current Research Topics:

• Learning Control

• Analysis and Control of Networked Systems

• Control Design of Sustainability Development

Learning Control

In many modern industries, the standard manufacturing process requires the repeated execution of the same series of tasks. For example, on a car making assembly line, each car will pass the same assembly procedures from the beginning to the end. Repeated executions of these tasks constitute the car manufacturing process. Hence, the quality and performance of each task execution is of crucial importance to the overall system, which will directly affect the cost and efficiency of the whole industry.

To achieve certain performance of each task execution, most current industrial systems use a fixed control strategy, that is, the control method is set up once and then remains unchanged when the system is running. This will result in satisfactory system performance in a lot of cases but when there exists disturbance or model changes/uncertainty, system performance may begin to deteriorate or even become totally unacceptable.

Iterative learning control (ILC) is a control method for improving performance of systems that operate in a repetitive manner. By learning information from previous executions of a task, ILC could use the repetition to achieve excellent system performance in terms of highly accurate (theoretically perfect) tracking of a desired signal. ILC has wide applications in manufacturing, chemical batch processing, robotics and some medical equipment. Since the original work of Arimoto in the mid 1980s, the general area of ILC has been the subject of intense research effort.

In this research, we aim to develop new ILC algorithms with better performance in terms of convergence and robustness. We are particularly interested in optimisation based design methods. We are also exploring their potential applications to mechanical testing and, more recently, to stroke rehabilitation and quadrotor UAVs.

Analysis and Control of Networked Systems

Complex networked systems have found tremendous applications in many areas, e.g. biological cellular network, mobile network, power system and internet social network. These systems usually consist of a vast number of components (agents), which, through interacting with each other, could achieve complex global goals. For example, diverse biological rhythms are generated by multiple cellular oscillators that manage to operate synchronously. To get a better understanding of collective behaviors of the whole network and to design control policies for local agents to achieve a global target, remains a grand challenge.

A more deep insight into these systems reveals that, in order to achieve a complex global goal, the agents exchange information about themselves with neighborhoods, and adapt their behavior (control action) according to the obtained information. By appropriately adjusting their behaviors, the agents are more ‘coordinated’ and move towards the direction of the desired global target.

Motivated by this observation, we attempt to address the important questions: how could we understand the structure, dynamics and collective behaviors of complex networked systems? Can we design control laws using only local information to achieve a complex global target?

Control Design of Sustainable Development

Climate change and global warming is one of the most important and pressing issues facing our planet. It is expected that if no action is taken, the temperature rise by year 2100 will be around 4oC (compared to pre-industrial levels), resulting in widespread loss of bio-diversity, decrease in agricultural productivity, significant rise of sea level (primarily from the melting of the Greenland ice sheet) and rising intensity of extreme weather events. Globally, annual emission now rise to 30.6 gigatonnes of CO2 per annum (according to IEA data), and this figure has no signs of going down.

In this research, we use concepts from modern systems control theory in sustainable policy design for climate change mitigation. We aim to develop algorithms for determining the optimal policy that both achieves sustainable levels of emissions of CO2 (and other greenhouse gases) and minimises the impact on the economy, but also explicitly addresses the high levels of uncertainty associated with predictions of future emissions and climate change. Including uncertainty within the design quantifies the risk associated with the emissions policy, which allows policy makers and emitters of CO2 to incorporate risk within their strategic plans.