Particle Swarm with Domain Partition and Control Assignment for Time-Optimal Maneuvers

What is it about?

A numerical approach for minimum-time maneuvers, based on the Particle Swarm Optimization, is proposed. The external control structure is assigned as a Bang-Bang control policy whilst the search space is divided into several sub-domains. Each sub-domain is defined by the number of switches per axis and by the sign of the bangs. The algorithm searches for the minimum-time maneuver changing the time instants associated with the switches of the bang-bang external control and moving from one sub-domain to another. The migration among the several sub-domains is assured by the introduction of the Push-In and Push-Out features which enhance the standard mobility of the particles in the swarm. For general non-linear problems, the maximum number of allowed switches is unknown, however it may be imposed based on some insight on the problem. Though the imposed bang-bang structure may lead to the presence of many sub-optimal solutions, the proposed approach is able to recognize the global minimizing solution for problems with end-point conditions and path constraints. Two different test cases are reported to validate the method by comparison with other results from the literature.

Featured Image

Why is it important?

Compared to other works by the authors, this paper proposes a direct dynamics approach based on the application of the particle swarm optimization technique. For minimum-time maneuvers of control-affine dynamical systems, where the necessary conditions of optimality are knwon, the particle swarm can be used to search for the optimal bang-bang policy. The results reported in this paper are interesting compared to other similar approaches proposed by other researchers in this field.

Perspectives