Computer Algebra and High-Performance Computing Support for Model Predictive ControlComputer Algebra and High-Performance Computing Support for Model Predictive Control

In many industrial and engineering applications, process control plays a central role. Among the possible control strategies, model predictive control (MPC), also called receding horizon control (RHC), stands out for its excellent ability to handle constraints. While MPC has been successfully applied to many industrial applications, further developments are limited when it becomes necessary to solve many large on-line optimization problems. To overcome this situation, parametric optimization is often used such that most of the computation burden is pushed to an off-line phase. In this project, we propose to develop algorithms and software tools based on symbolic computation, to perform parametric optimization together with the corresponding on-line procedures, targeting MPC in the case of polynomial constraints and polynomial objective function. The proposed research will bring together the latest advances in computational real algebraic geometry and high-performance computing techniques. Our goal is to generate application driven packages for MAPLE, the flagship product of our industrial partner. In particular, the deployed infrastructure will be used to support another application relying on an off-line on-line strategy, namely code generation of parallel programs. Not only will this allow us to capitalize on the effort invested to support MPC via symbolic parametric optimization, but it will also serve as a development tool in this project by generating portable and efficient parallel code in support of our MPC solvers.

Faculty Supervisor:

Dr. Marc Moreno Maza

Student:

Changbo Chen, Parisa Alvandi, Ning Xie & Farnam Mansouri

Partner:

MapleSoft Inc.

Discipline:

Computer science

Sector:

Information and communications technologies

University:

Western University