The objective of this minisymposia is to gather specialists of numerical optimal
control (mostly of ordinary differential equations) as well as developers of state of the art softwares
dedicated to this topic. We focus on (i) direct methods that transcribe control problems into nonlinear
programs, and (ii) on indirect algorithms that leverage necessary optimality conditions before
discretising these problems. Presentations will in particular showcase recent developments based on
WORHP, CasADI and control-toolbox. Our goal is to go beyond mere comparisons and to
emphasise how these various approaches can be made to collaborate so as to get the best of several
worlds. This clearly includes chaining direct and indirect methods to capture the solution structure
(number and type of control arcs), then solve very accurately using a tailored shooting function. The
MS will also try to address related and currently very active topics such as automatic differentiation
(aka differentiable programming) in modern languages for scientific programming.
Mini symposium organizers:
Jean-Baptiste Caillau (Université Côte d’Azur)
Joseph Gergaud (Université de Toulouse, INP-ENSEEIHT-IRIT, CNRS, France)
Session 6. Room A4, Friday 09:00-11:00.
Chair: Joseph Gergaud (Université de Toulouse, INP-ENSEEIHT-IRIT, CNRS, France)
Speakers:
Matthias Knauer (Universität Bremen) WORHP Lab: teaching and showcasing numerical methods of optimization and optimal control
Jean-Baptiste Caillau (Université Côte d’Azur) Control-toolbox : Solving control problems within Julia
Dominik H. Cebulla (Institute for Mathematical Optimization, Technische Universität Braunschweig) Solving mixed-integer optimal control problems with a discretize-then-optimize approach using CasADi with an application to preparative chromatography
Paul Malisani (IFP Energies nouvelles, Applied Mathematics Dpt, Rueil-Malmaison, France) Optimize-then-discretize interior-point methods in optimal control: convergence results and primal-dual implementation