Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.


System Identification

Welcome to the Embedded Optimization Group



Embedded optimization in the loop Optimization plays a key role in enginneering, and in particular in the field of control. While traditionally it was regarded as an analysis and design tool, there has been significant research effort in the last decade to enable numerical optimization be implemented on resource constrained embedded platforms for optimal constrained decision making, which requires the solution of optimization problems at high rates in real-time. One of the landmark achievements in the field from this group was the solution of quadratic programs (QPs) online at Megahertz rates with application to atomic force microscopy.


What we do

The Embedded Optimization group at IfA develops algorithms, methods and software tools for numerical optimization, and works on applications that involve optimization as a tool for optimal decision making under constraints. In particular, we focus on embedded optimization for predictive control, system identification, parameter learning and moving horizon estimation, although much of the theory and algorithms developed can be applied in other problem settings as well.


Example: Customized solvers for predictive control

In this video, you see a comparison of two solvers when simulating a race between two 1:43 scale RC race cars that are controlled by model predictive control, maximizing their progress while staying on the track and avoiding obstacles.

On the left, the predictive control problems are solved by the state-of-the-art solver IBM CPLEX, while on the right a custom solver generated by FORCES is used. On average, the speedup is about a factor of 40x, which makes not only simulating (and tuning) much more convenient, but it also enables us to actually run the constrained optimal controller in real-time on an embedded platform. Make sure to also visit our Software page for more details.


Related Groups and Pages