PRE-FLIGHT: AI-Driven Orbital Trajectory Optimization

At Valar Space we believe space should be easy to fly. That vision pushed us to launch PRE-FLIGHT – our intelligent platform for orbital-trajectory optimisation. Over the past ten months we turned that idea into a working, TRL-6 solution that is already improving how satellites plan, manoeuvre and stay safe in crowded skies. Here is a behind-the-scenes look at the project: its purpose, the work we carried out and the milestones we have already reached.

Why we built PRE-FLIGHT

Operators of small-to-large satellite fleets face three converging challenges:

1. Ever-denser orbits. Collisions and costly avoidance burns are becoming routine.
2. Low-thrust propulsion. Electric engines need smarter, longer-horizon planning.
3. Pressure to automate. Tomorrow’s constellations run on software, not 24/7 control rooms.

PRE-FLIGHT answers those needs. It couples the Q-law low-thrust guidance method with modern machine-learning. A neural network learns the best weightings for Q-law in advance, then produces near-instant recommendations on-orbit, cutting on-board CPU time and propellant at the same time.

How we turned theory into software

PhaseMain activitiesOutcomeSystem & data-pipeline buildCreated a high-fidelity simulation core, synthetic-data generators and connectors to external SSA providers (ExoAnalytic, Spaceflux, COMSPOC).Continuous stream of labelled trajectories for ML training.ML model design & trainingTrained neural nets on millions of simulated transfers; refined with heuristic genetic algorithms and human-in-the-loop review.Model predicts optimal Q-law weights in <50 ms.Integration & cloud deploymentWrapped the engine behind a REST-first API and an operator-friendly dashboard; hardened under Spring Boot micro-services with websocket push updates.“Flight-Dynamics-as-a-Service” accessible from any ground segment.Pilot campaignsGravity Space, Es’hailSat, Infinite Orbits, Space Inventor & Astranis ran end-to-end trials covering GEO station-keeping, relocation burns and collision avoidance.Validated ∆V savings up to 12 % and planning time 20× faster than manual ops.

All development artefacts, test plans and review notes live in a private GitHub monorepo so every line of code remains traceable and extensible.

Achievements we’re proud of

• TRL-6 demonstration – The complete stack has been exercised in customer-relevant scenarios, proving readiness for operational uptake in 2025.
• Autonomous decision support – PRE-FLIGHT recommends manoeuvres that comply with operator constraints and traffic-management rules, then feeds them straight into our Valar FDaaS platform.
• Scalability baked in – Cloud-native architecture scales from a single cubesat to multi-satellite constellations without extra licences or hardware.
• Early market traction – Seven paying pilot contracts secured and counting, spanning Europe, North America and the Middle East.

What’s next

Our 2025 roadmap focuses on three fronts:

1. Live-ops rollout – Transition pilots to continuous service, expanding to LEO constellation customers.
2. Regulatory data-sharing – Integrate forthcoming STM data-schemas to help agencies visualise and audit manoeuvre intentions in real time.
3. Research & community – Publish our findings in Acta Astronautica and present at AAS/AIAA, ISSFD and IAC, because open dialogue drives safer orbits for everyone.

Closing thought

Space is getting busy, but with PRE-FLIGHT on board operators can think less about maths and more about mission. We are thrilled by the early results and grateful to our pilot partners for helping shape a tool that keeps humanity’s highways above Earth open, efficient and safe.

Keep watching this space – literally – for more updates as PRE-FLIGHT goes operational. 🚀

‍