Picture of aircraft jet engine

Strathclyde research that powers aerospace engineering...

The Strathprints institutional repository is a digital archive of University of Strathclyde's Open Access research outputs. Strathprints provides access to thousands of Open Access research papers by University of Strathclyde researchers, including by Strathclyde researchers involved in aerospace engineering and from the Advanced Space Concepts Laboratory - but also other internationally significant research from within the Department of Mechanical & Aerospace Engineering. Discover why Strathclyde is powering international aerospace research...

Strathprints also exposes world leading research from the Faculties of Science, Engineering, Humanities & Social Sciences, and from the Strathclyde Business School.

Discover more...

Global optimisation of multiple gravity assist trajectories

Ceriotti, M. (2010) Global optimisation of multiple gravity assist trajectories. PhD thesis, University of Glasgow.

[img] PDF
Ceriotti_M_strathprints_Global_optimisation_of_multiple_gravity_assist_trajectories_PhD_Thesis_2010.pdf - Preprint

Download (6MB)


Multiple gravity assist (MGA) trajectories represent a particular class of space trajectories in which a spacecraft exploits the encounter with one or more celestial bodies to change its velocity vector; they have been essential to reach high Delta-v targets with low propellant consumption. The search for optimal transfer trajectories can be formulated as a mixed combinatorial-continuous global optimisation problem; however, it is known that the problem is difficult to solve, especially if deep space manoeuvres (DSM) are considered. This thesis addresses the automatic design of MGA trajectories through global search techniques, in answer to the requirements of having a large number of mission options in a short time, during the preliminary design phase. Two different approaches are presented. The first is a two-level approach: a number of feasible planetary sequences are initially generated; then, for each one, families of the MGA trajectories are built incrementally. The whole transfer is decomposed into sub-problems of smaller dimension and complexity, and the trajectory is progressively composed by solving one problem after the other. At each incremental step, a stochastic search identifies sets of feasible solutions: this region is preserved, while the rest of the search space is pruned out. The process iterates by adding one planet-to-planet leg at a time and pruning the unfeasible portion of the solution space. Therefore, when another leg is added to the trajectory, only the feasible set for the previous leg is considered and the search space is reduced. It is shown, through comparative tests, how the proposed incremental search performs an effective pruning of the search space, providing families of optimal solutions with a lower computational cost than a non-incremental approach. Known deterministic and stochastic methods are used for the comparison. The algorithm is applied to real MGA case studies, including the ESA missions BepiColombo and Laplace. The second approach performs an integrated search for the planetary sequence and the associated trajectories. The complete design of an MGA trajectory is formulated as an autonomous planning and scheduling problem. The resulting scheduled plan provides the planetary sequence for a MGA trajectory and a good estimation of the optimality of the associated trajectories. For each departure date, a full tree of possible transfers from departure to destination is generated. An algorithm inspired by Ant Colony Optimization (ACO) is devised to explore the space of possible plans. The ants explore the tree from departure to destination, adding one node at a time, using a probability function to select one of the feasible directions. Unlike standard ACO, a taboo-based heuristics prevents ants from re-exploring the same solutions. This approach is applied to the design of optimal transfers to Saturn (inspired by Cassini) and to Mercury, and it demonstrated to be very competitive against known traditional stochastic population-based techniques.