Intelligent Systems

My research interests can be defined as the interface between control engineering, artificial intelligence and machine learning that includes intelligent systems and soft computing concepts. The research theme can be grouped under the broad heading of ‘goal-based learning control’. This encompasses a wide range of technologies from intelligent search-based strategies such as learning automata and evolutionary based techniques (genetic algorithms, genetic programming) to reinforcement learning techniques, such as Q-learning and adaptive heuristic critic designs. I am particularly interested in the development and extension of these techniques and methodologies to provide practical control solutions to real engineering applications. These methods should be robust, stable and provide benefits in terms of greater system efficiency and flexibility. They should also provide solutions that can be implemented on-line in real-time embedded control environments and that are suitable for continuous operation, responding to both changing conditions and changing system demands.

My research examines goal direct reinforcement learning from a control engineering perspective. Taking ideas from areas such as optimal control and applying learning methodologies to provide practical tuning solutions. I have previously focused on Genetic Algorithms and Learning Automata based approaches. This has resulted in a new algorithm the Genetic Learning Automata that combines the genetic algorithms search potential with the discrete learning automata. This provides a stopping rule and an indication of how well the genetic search is progressing. This has also been cited by researchers in Finland who have used it for the control of combustion.

I have extended the discrete learning automata to continuous actions with a new algorithm (CARLA: Continuous Action Reinforcement Learning Automata). This approach adapts a probability density function that is used to select the best action in stochastic environments. I have applied this to learning active and semiactive suspension control and for engine idle speed control. It has also been used by other researchers in asynchronous electric motor control, active magnetic bearing control and global training of hidden Markov models. I have recently expanded the CARLA algorithm to multi-objective optimization and a distributed hardware (multiple-Arduinos) implementation. I am also planning to publish a book on engineering applications of learning automata.

Control Engineering

My interest in control systems engineering started when I first studies control technology at age 16. I find it extremely rewarding to apply theoretical concepts to successfully control complex industrial problems.

My principle research interest which relates to the intelligent systems research above is the developing of practical control system tuning solutions for industrial control systems.

Diagnostics, Prognostics & Vehicle Health Management

The increasing complexity of today’s automobiles raises the importance of maintaining the vehicles health and of correctly identifying faults and imminent system failures. Advanced diagnostics algorithms to isolate the source of a problem, prognostics algorithms determine when components, such as batteries and brake-pad replacement is required and fault tolerant control mitigates problem severity and minimizes the impact on the driver enabling the convenient scheduling of maintenance Model based, data-driven diagnostics and prognostics systems can provide advanced information to maintain system operation in the event of fault or error conditions. This can be combined with a controller redundancy to provide a framework for enhanced vehicle health. The vision outlined above still requires a large amount of research work.

Automotive Systems

Automotive systems offer the perfect platform for the development of advanced control algorithms, new diagnostics, prognostics and vehicle health management solutions, and intelligent systems technology. Today’s vehicles contain a hybrid powertrain, an integrated distributed electrical architecture, advanced vehicle telematics systems and controllable chassis technologies (semi-active suspension, active front steering etc). This actuation together with inertial vehicle sensors, radar and cameras can extent vehicle safety systems from stability control to collision avoidance. The technology can also be applied beyond the individual vehicle with applications exploiting ‘vehicles to vehicle’ and ‘vehicle to infrastructure’. There are many examples of complex systems from different interacting vehicle subsystems to vehicle interactions while driving. The automotive arena provides a real test environment for integrated diagnostics, prognostics and control systems at many different levels. The potential applications range from single component diagnostics and subsystem control to integrated multi-vehicle platoon control systems.