A Parallelization Strategy for Discrete Variational Problems


Sebastián Ferraro from Universidad Nacional del Sur & CONICET presents research on discrete variational methods, which have showcased exemplary performance in numerical simulations of mechanical systems. The presentation elaborates on an iterative procedure tailored for the resolution of variational equations with boundary conditions. These equations correspond to arbitrary order discrete Lagrangians. Central to this procedure is a parallelization strategy that taps into the potentials of GPUs (graphics cards) or multicore CPUs.

As an application showcase, Sebastián will discuss the Zermelo navigation problem and a fuel optimization issue in a controlled four-body problem in astrodynamics. The talk will also touch upon conditions necessary for the convergence of this method.

This session is rooted in joint works with David Martín de Diego and Rodrigo Takuro Sato Martín de Almagro, both researchers under this project funded by BBVA Foundation and Agencia Estatal de Investigación.