Year: 2011
Type: conference-paper
Abstract: At the European Space Agency a software framework was developed over the last few years focusing on global optimization. The framework, mainly developed in-house at the Advanced Concepts Team (ACT), was released as an open-source project last year, that allows to run high dimensional global optimization problems, like interplanetary trajectory planning and optimizing controllers for space robotics. PaGMO is implemented in an object-oriented fashion as a C++ library but also has a Python wrapper for easy interactive sessions [1]. The core library includes support for multi-core systems (mainly through multi-threading) and recently has been trying to add functionality towards cluster and networked computers by aiming for an MPI integration. The project described in this paper was published as a possible Google Summer of Code project on the PaGMO sourceforge page but was dropped because of limited funding. Over the last year the aim was set to implement capabilities to run the PaGMO optimization heuristics on GPU architectures. As a starting point the evolving docking problem was chosen, which aims to design a neuro-controller for the automatic rendezvous and docking task in spacecraft operations. A genetic algorithm is used to search for a neural network that controls the spacecraft [2]. The docking process involves a series of maneuvers and controlled thruster burns, which lead to the chaser vehicle eventually being able to come close and finally attach to the, in our case passive, target spacecraft. It can be seen as a variant of the phototaxis task in robotics, but using a different physical environment (no friction, non-zero external forces, thruster-based motion). The spacecraft is modeled as a circle with thrusters at the edge of its size (radius), which control forward/backward and rotational acceleration.
Authors: M. Kashif Kaleem, Jürgen Leitner
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
Funder: European Space Agency
OpenAlex ID: W2610359274