Background
James Fleming is a Lecturer within the Wolfson School of Mechanical, Electrical and Manufacturing Engineering at 天堂视频, joining the school in September 2019.
James obtained the MEng and DPhil degrees in 2012 and 2016 respectively from the University of Oxford, where he studied control engineering and developed algorithms for Model Predictive Control of uncertain state-space systems as part of his doctoral research. From 2016 to 2019 he was a Research Fellow at the University of Southampton, developing driver models and optimal control algorithms for the G-Active (Green, Adaptive ConTrol of Interconnected VEhicles) project, which used knowledge of driver preferences to save fuel and reduce emissions in the energy management of conventional, hybrid and electric vehicles.
Qualifications
- 2012—2016 University of Oxford, Department of Engineering Science, Doctor of Philosophy in Engineering Science
- 2008 – 2012 University of Oxford, St Edmund Hall, Master of Engineering Science
Optimal control of constrained and uncertain systems
Many important control problems in industry involve optimising a performance measure such as cost or energy consumption of a process while satisfying constraints, which may be safety critical. But real-world control systems must deal with uncertainty in the controlled process, giving satisfaction of constraints and good performance in all or almost all cases.
Advanced modelling and control techniques have great potential to improve energy economy and safety in the automotive and renewable energy sectors. This is especially true given the recent pushes towards electrification of powertrains and incorporation of driver assistance systems in cars, with all new cars in the UK planned to be zero emission by 2040, and several forms of safety-related driver assistance such as autonomous emergency braking, lane-keeping assistance, and intelligent speed assistance being mandatory in new vehicles within the EU from 2022. In the renewable energy sector, advanced control can improve the lifespan and output of wind turbines and wind farms.
Model predictive control and its applications
Model Predictive Control (MPC) forms a powerful framework for constrained control in which the state of the system is predicted and optimised over some future horizon, but work is still ongoing to develop algorithms that account for uncertainty. Naive approaches typically have a computational complexity that grows exponentially as predictions are made further into the future, making them unusable. To remedy this, we have developed approaches that consider sets of predicted states rather than the states themselves. This adds a little conservatism but leads to controllers that work very well in practice, with many applications throughout the automotive, electrical power and renewable energy industries.
Some recent examples are the design of driver assistance and autonomous driving systems that use modelling and prediction of conventional and electrified powertrains and upcoming traffic conditions to reduce energy usage when following other vehicles, intelligent vehicle air conditioning that uses less fuel by scheduling use of a compressor when the combustion engine is operating at high efficiency, and gyroscopic stabilisation systems for motorcycles that enhance stability and handling during cornering and under heavy braking.
Module leader for:
- WSB045, Electrical Power and Machines
- WSC104, Robotics and Control
Lecturer on:
- TTD108 / WSD522 / WSP022 - Power Electronics, Machines and Drives
- WSD527 / WSP027 - Advanced Methods for Control