Search

Modern Control: making optimal decisions online

William Mong Distinguished Lecture by Professor Jan M. Maciejowski
Nov 28, 2016

Professor Jan M. Maciejowski, Professor of Control Engineering at Cambridge University, gave a lecture on November 28, 2016 titled “Modern Control: making optimal decisions online”.

 

The most successful and widely used ‘advanced control’ paradigm that has emerged since about 1980 is Model Predictive Control (MPC). This took full advantage of modern computing power, and of modern optimization algorithms, to determine control signals by solving optimal control problems online in ‘real time’, on the basis of the most recent measurements. MPC was initially used for the control of petrochemical plants, which did not require such problems to be solved very quickly. It can now be used for the control of a very wide range of processes and devices, from power electronics which require MHz solution rates, to extremely complicated non-convex problems such as air-traffic control and enterprise planning that would once have been thought of as ‘management’ rather than ‘engineering’ problems. The talk described these and other applications, such as fault-tolerant flight control and spacecraft control, and described the various implementation technologies appropriate to each kind of problem, including special-purpose hardware such as FPGAs and GPUs. The talk also pointed out that, while very powerful, optimisation is not always an easy-to-use panacea. The talk aimed at a general engineering and scientific audience.