TrajGPT: Beyond Low Earth Orbit Trajectory Generation via Machine Learning

Ensuring a deep space crew can arrive at their destination at the appropriate time and return safely requires detailed planning. Current design methods involve generating millions of trajectories and analyzing countless permutations of variables such as launch date and time, propellant budgets, and abort options. This is too complex to be done onboard in real-time; however, changing the modeling techniques may enable onboard capabilities that could improve crew autonomy and operational flexibility. In year one, the TrajGPT IRAD project investigated use of Machine Learning (ML) to possibly replace the current brute-force methods of physics models. The results to date have shown promise and warrant further investigation. In year two, the model will be trained on additional scenarios and results will be validated against current ground software. Year two objectives include demonstrating generation of 1) no-fail ascent trajectories every 1 minute through a launch window and 2) on-orbit abort trajectories at various points in a lunar mission.

The general capability could be broadly applicable to both crewed and uncrewed missions involving spaceflight. ​For crewed missions, a change in modeling techniques may enable onboard capabilities that could improve crew autonomy and operational flexibility, vital to the success of future Mars missions.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.