|
Thursday, 12 July 2012 |
|
|
MSc presentation:
|
Fast optimization-based control and estimation using operator splitting methods
|
|
|
|
Speaker:
|
Georgios Stathopoulos
|
|
|
|
Supervisor:
|
Prof.dr.ir. B. De Schutter, dr. T. Keviczky
|
|
|
|
Location:
|
room L
|
|
|
|
Time:
|
10:00 until 10:45
|
|
|
|
Abstract:
|
The size of modern systems has grown significantly and the need for fast, online optimization-based control is a necessity. Exploiting the modern, mul- ticore architectures we can achieve fast implementations by splitting a large problem into several smaller ones that can be solved in parallel. In this thesis, we apply an operator splitting technique to a generic convex optimal control problem. The resulting algorithm alternates between a quadratic regulator it- eration, and a step in which single-period optimization problems are solved in parallel. Depending on the constraints and non-quadratic objective terms, these single-period problems can be solved extremely quickly, or even analytically in many cases. In the time-invariant case, pre-computing the gain matrices in the quadratic regulator problem gives another speed-up, as well as an algo- rithm that requires no division, and is therefore suitable for implementation with fixed-point processor. The method is demonstrated on several examples arising in different application areas. Furthermore, we develop an extension to this algorithm so that it can handle more generic, optimal control problems as well as a different decomposition approach that achieves even more paralleliza- tion than the originally proposed algorithm.
|
|
