BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Code Generation for Fast Embedded Optimization
DTSTART:20120529T110000
DTEND:20120529T121500
DTSTAMP:20260407T175815Z
UID:83c8ff47212a12a28bc940a6e2c237020774b6e74a91980a22497a57
CATEGORIES:Conferences - Seminars
DESCRIPTION:Alexander Domahidi and Stefan Richter\, Automatic Control Labo
 rators\, ETHZ\, Zürich\, Switzerland\nIn this talk\, we discuss two code 
 generation approaches to solving convex optimization problems\, occurring 
 e.g. in model predictive control (MPC)\, efficiently on embedded platforms
 .\n\nIn the first part of the talk\, we introduce FiOrdOs. This toolbox co
 nsiders code generation for the class of multi-parametric convex programs 
 with a quadratic cost and a feasible set given as the intersection of an a
 ffine set and a `simple' convex set for which a projection can be evaluate
 d at low cost. The toolbox implements both polyhedral and non-polyhedral s
 imple sets\, e.g. the simplex and the second-order cone respectively. Thus
 \, solver code for problems beyond quadratic programming can be generated.
  The implemented solution approaches either use the gradient or the fast g
 radient method in the primal domain or resort to Lagrange relaxation if eq
 uality constraints are present. Additional toolbox features include optima
 l preconditioning and the automatic certification of the iteration count f
 or a restricted set of problems. The generated C-code can be compiled for 
 any platform and can be made library-free. FiOrdOs also provides a tailore
 d MEX-interface for calling the generated solvers inside Matlab and a Simu
 link library for rapid prototyping. In the talk\, we will recap first-orde
 r methods and illustrate the features of FiOrdOs on an AC/DC converter con
 trol problem.\n\nIn the second part of the talk\, we discuss efficient int
 erior point methods that are tailored to convex multistage problems (which
  MPC is a special case of)\, and specify important algorithmic details req
 uired for a high speed implementation with superior numerical stability. I
 n particular\, the presented approach allows for quadratic constraints\, w
 hich is not supported by existing fast solver implementations. We present 
 numerical studies obtained with FORCES\, a tool which is implementing the 
 proposed methods in a code generation framework\, and compare our solver t
 o three well-known solver packages CPLEX\, OOQP and CVXGEN\, outperforming
  the fastest of these by a factor 2-5 in speed and 3-70 in code size. More
 over\, our solver is shown to be very efficient for large problem sizes an
 d for quadratically constrained QPs\, extending the set of systems amenabl
 e to advanced MPC formulations on low-cost embedded hardware.
LOCATION:ME C2 405 http://plan.epfl.ch/?room=ME%20C2%20405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR