Parallel Iterative Methods for Variational Integration and Related Problems


Rodrigo Takuro Sato Martín de Almagro from the Institute of Applied Dynamics at Friedrich-Alexander-Universität Erlangen-Nürnberg presents our research on discrete variational methods, which have demonstrated outstanding performance in numerical simulations of various mechanical systems. In this talk, an iterative procedure is introduced to solve discrete variational equations for boundary value problems. Specifically, the discussion delves into a parallelization strategy optimized for multicore CPUs and GPUs.

The focus then shifts to the application of this parallel method for higher-order Lagrangian systems, encountered in fully-actuated problems and more. Analyzing the convergence conditions of these techniques has ushered us into exploring the discrete Jacobi equation, among other intriguing challenges.

This presentation is a collaborative effort with Sebastián J. Ferraro from Universidad Nacional del Sur & CONICET and David Martín de Diego from Instituto de Ciencias Matemáticas.

The research for this project is funded by BBVA Foundation and Agencia Estatal de Investigación.