Interplanetary Trajectory Optimisation using Particle Swarm Optimisation

Global heuristic optimisation and Monte Carlo dispersion analysis for Earth-to-asteroid intercept trajectories, presented at JAXA

Interplanetary Trajectory Optimisation using Particle Swarm Optimisation

Presented at: 35th JAXA Astrodynamics Workshop (ISAS / JAXA)  |  Paper record: jaxa.repo.nii.ac.jp/records/2002502
Methods: particle swarm optimisation (PSO), SciPy, two-body numerical propagation, Monte Carlo dispersion

Open Conference Paper (PDF) Download PDF JAXA Repository Record Jump to Document Viewer

Summary

Finding fuel-efficient transfers to small bodies is a hard optimisation problem. Targets such as 3200 Phaethon (the target of JAXA’s DESTINY+ mission, with eccentricity $e = 0.89$ and inclination $i = 22.3^\circ$) and the dwarf planet 1 Ceres combine nonlinear dynamics with tight $\Delta V$ budgets. Gradient-based solvers tend to get stuck in local minima, and grid searches over launch date and flight time are slow.

I wrote a trajectory optimisation pipeline in Python based on particle swarm optimisation (PSO). A swarm of candidate trajectories (“particles”) explores the search space, and the particles share information so the swarm moves towards the departure velocity and flight time with the lowest $\Delta V$. Orbits are propagated numerically, alongside an analytical Kepler solver that handles high eccentricities and converges in under a second.

Results

  • A direct Earth-to-Phaethon transfer taking 212.7 days.
  • The solver converged within 40 to 60 iterations with 30–50 particles, without a hand-picked initial guess.
  • Viable multi-target sequences (Earth $\to$ 3200 Phaethon $\to$ 1 Ceres), checked for consistent arrival and departure velocities and orbital phasing.
  • A 500-run Monte Carlo dispersion analysis with thruster execution errors ($\pm 3\sigma$) and pointing errors. Without correction only 3.4% of flights reached the target, which sets the margins and tracking cadence needed for mid-course trajectory correction manoeuvres (TCMs).

Trajectories and convergence

Figure 1: 3D heliocentric trajectory showing Earth departure, intercept of asteroid 3200 Phaethon, and subsequent transfer to dwarf planet 1 Ceres.
Figure 2: 2D ecliptic plane projection showing orbital phasing.
Figure 3: Global best objective score convergence across swarm generations.

Conference paper

The paper from the 35th JAXA Astrodynamics Workshop is below.