gmat-mission-design

GMAT mission design

You have a real astrodynamics simulator in the loop. The core discipline: never cite an orbital-mechanics number you computed in your head — cite numbers from a GMAT report. Your analytic estimates are the acceptance test for the simulation, not the deliverable.

The loop

1. Read the idioms first. Call getGmatIdioms before writing any script. It is a curated list of gotchas that each cost a failed run to discover (cumulative ElapsedSecs, ASCII-only scripts, conditional syntax limits, finite-burn multi-spacecraft rules...). Skipping this step reliably costs more calls than it saves.

2. Seed from a sample when one matches. listGmatSamples / getGmatSample give you NASA's known-good scripts (lunar transfer, Mars B-plane, electric propulsion, finite-burn targeting, station-keeping...). Adapt rather than invent for anything beyond basic LEO work. When adapting for headless runs: strip OrbitView/XYPlot subscribers, keep the force models and targeter structure, add a ReportFile.

3. Write down analytic expectations before running. Vis-viva, Hohmann ΔVs, phasing-orbit periods, plane-change costs, drag ballpark from scale heights. Two reasons: they seed your Vary initial guesses and bounds, and they are how you'll know the run is right and not merely converged.

4. Write the script. Pure ASCII (no accented characters, not even in comments). One statement per line. Create a ReportFile with WriteHeaders = false and Report every number you intend to cite. Output lands in GMAT's output/ dir and comes back in the reports field of the result.

5. Run with runGmat and branch on stage:

   • parse — syntax/object error; the message names the line. Re-check against the idioms.
   • convergence — if a Vary variable is pinned at a bound with large constant variance, the design is infeasible (rethink goals/bounds); otherwise widen bounds, loosen tolerance past physical oscillations, or pick a less degenerate goal.
   • run — propagation/stop-condition failure. Classic causes: stop on an apsis of a (near-)circular orbit, stop condition already satisfied (cumulative ElapsedSecs).
   • timeout — raise timeoutSec only when the propagation is genuinely long (multi-day low-thrust spirals); otherwise suspect an infinite loop and check raw_tail to see where it hung.