Geometric numerical methods for Lie systems and their application in optimal control

TL;DR

This paper introduces geometric numerical methods, including Magnus and Runge-Kutta-Munthe-Kaas, for solving Lie systems, with applications to optimal control problems in mechanics, demonstrating their accuracy through examples.

Contribution

It develops and adapts geometric numerical methods specifically for Lie systems, enhancing analytical and numerical integration techniques.

Findings

Methods accurately solve Lie systems on SL(n,R)

Numerical solutions effectively applied to vehicle control problem

Enhanced understanding of geometric integration for Lie systems

Abstract

A Lie system is a non-autonomous system of first-order ordinary differential equations whose general solution can be written via an autonomous function, a so-called (nonlinear) superposition rule of a finite number of particular solutions and some parameters to be related to initial conditions. Even if the superposition rules for some Lie systems are known, the explicit analytic expression of their solutions frequently is not. This is why this article focuses on a novel geometric attempt to integrate Lie systems analytically and numerically. We focus on two families of methods: those based on Magnus expansions and the Runge-Kutta-Munthe-Kaas method, which are here adapted to the geometric properties of Lie systems. To illustrate the accuracy of our techniques we propose examples based on the SL$(n,\mathbb{R})$ Lie group, which plays a very relevant role in mechanics. In particular, we depict an optimal control problem for a vehicle with quadratic cost function. Particular numerical solutions of the studied examples are given.