Speaker
Dr
Alexander Wittig
(ESA Advanced Concepts Team)
Description
The optimization of low-thrust trajectories is a difficult task. While techniques such as Sims-Flanagan transcription gives good results for short transfer arcs with at most a few revolutions, solving the low-thrust problem for orbits with large numbers of revolutions is much more difficult.
We developed a massively parallel genetic optimization algorithm to obtain low-thrust solutions for targeting of a celestial body, such as the Moon or a planet. The solution is not limited to simple arcs or few rotations, but is capable of solving the problem also in the case of many revolutions where other classical methods fail. While we perform the propagation in a two-body model, in principle the propagation can also be performed in more complete models such as the circular restricted three body problem.
The optimization algorithm chosen is a genetic algorithm with large population size. Due to its massively parallel nature, this type of problem is a natural fit for implementation on a GPU. Modern GPUs are capable of running thousands of computation threads in parallel, allowing for very efficient evaluation of the fitness function over a large population. In particular, we optimize the shape of the control function as well as the departure time. Propagation of the spacecraft is then performed in massively parallel fashion on the GPU before the results of the fitness function are read back into CPU memory for preparation of the next iteration of the algorithm.
We demonstrate with various examples how this algorithm can provide good initial guesses for a following local optimization to compute very accurate low-thrust orbits.
| Applicant type | First author |
|---|
Primary author
Dr
Alexander Wittig
(ESA Advanced Concepts Team)
Co-author
Dr
Dario Izzo
(ESA ACT)
