- Disciplines
- Laboratory
- SIGNALS AND SYSTEMS LABORATORY
- Host institution
- CentraleSupélec
- Doctoral school
- Information and Communication Technologies (STIC) – ED 580
Description
Many approaches for the design of control laws rely on the solution to optimization problems (H2 design, predictive control, robust control, controller retuning…). In the case if large-scale, nonlinear systems, with possibly numerous integre variables (hybrid systems), these constrained optimization problems become difficult to solve and untractable. In addition, most of well known criterion (overshoot, time response, stability margins, norms…) appear to be non differentiable or even non analytic.
The orientation towards approximated and often stochastic methods, namely the metaheuristics (genetic algorithms, ant colonies, swarm optimization to cite just the most known ones) appears as an interesting alternative for the fast determination of satisfying solutions to difficult optimization problems, and so is a way to compute some control laws or controller retuning. Past works of the team [1], [2], [3] proved the interest of the idea. However, in the case of Automatic control, the stochastic behavior of these methods may bring some problems in terms of performance and robustness, for instance if these methods have to be used in an online procedure : without any guarantee on the solution obtained by these non repeatable algorithms, how can we guarantee the stability of the control law as well as its performances ?
Skills required
The candidate should have a good expertise in control (simple control laws, Hinfinity design) and knowledge inoptimization. The candidate will have to implement his / her algorithms and control laws on simulators and lab Equipment. Good skilss in programming are thus required.
Bibliography
[1] G. SANDOU. Metaheuristic optimization for the design of automatic control laws. ISTE-Wiley, 144 pages, Collection FOCUS Automation and control series, ISBN: 978-1-84821-590-0, 2013.
[2] Philippe Feyel. Optimisation des correcteurs par les métaheuristiques Application à la stabilisation inertielle de ligne de visée. Automatique / Robotique. CentraleSupélec, 2015. Thèse de doctorat
[3] Feyel, Philippe and Duc, Gilles and Sandou, Guillaume. Fixed-Order H-infinity Loop-Shaping Synthesis: a Time Domain Approach, IEEE SSCI, Singapore, 2013.
[4] Feyel, Philippe and Duc, Gilles and Sandou, Guillaume. Evolutionary fixed-structure Hinfinity synthesis for multiple plants, IEEE MSC, Antibes, France, 2014.
[5] I. Maruta, T-H. Kim and T. Sugie, Synthesis of fixed-structure H controllers via constrained Particle Swarm Optimization, 17th IFAC World Congress, Seoul, Korea, July 2008, pp 7843-7848.
Keywords
Metaheuristics, Optimization, Control law performances, Robustness
Application deadline 12/05/17
Duration36 months
Start date01/10/17
Creation date27/04/17