I wanted to model a rocket mission as one connected engineering decision, not a handful of scripts that stop at a target apogee.
← Back to all work
Engineering case study · Aerospace & Autonomy · 2026
Apogee — Rocket Mission Simulation & Verification
A systems-engineering pipeline built with RocketPy, Astropy, NumPy/SciPy, and Pydantic for rocket mission design, flight simulation, uncertainty analysis, recovery trade studies, and flight-readiness reporting.
Central idea
What this system investigates.
A useful mission-analysis tool should connect design decisions, simulation, uncertainty, trade-offs, and verification instead of treating them as isolated calculations.
Problem
Why this system exists.
Rocket mission analysis often becomes a collection of disconnected scripts: one for motors, another for stability, another for dispersion, and another for recovery. Apogee connects these stages through typed data contracts so assumptions, constraints, uncertainty, and downstream effects remain visible across the workflow.
System architecture
How the system is organized.
- Environment & launch window
- Motor selection
- Stability vs Mach
- 6-DOF Monte Carlo
- Recovery Pareto study
- Flight-data comparison
- Readiness report
Implementation
Methods and engineering choices.
The pipeline moves from launch-environment checks through motor selection, Mach-dependent stability analysis, six-degree-of-freedom Monte Carlo dispersion, recovery optimization, confirmation analysis, optional flight-log comparison, and automated reporting. RocketPy provides the flight-simulation foundation while Pydantic contracts make stage interfaces explicit and ProcessPoolExecutor parallelizes independent dispersion runs.
Results and evidence
What the current record demonstrates.
- Public v1.0 systems-engineering pipeline integrating RocketPy with Astropy, NumPy/SciPy, and Pydantic.
- 120-run six-degree-of-freedom Monte Carlo demonstration with seeded uncertainty inputs across wind, thrust, mass, CG, and launch conditions.
- Demo configuration records 2,200 m target apogee, K740-Synth selection, 0.12% apogee error, and 1.96 cal minimum static margin.
- Recovery optimization uses a Pareto front for horizontal drift and impact velocity, followed by a full uncertainty-aware confirmation run.
- 19 unit tests cover motor parsing, impulse classification, recovery Pareto logic, and stability-selection behavior.
- Synthetic motor and flight-log inputs are explicitly labeled; the repository does not claim physical validation or flight certification.
Verification
Checks performed and their scope.
- 19 unit tests cover motor parsing, impulse classification, recovery Pareto logic, and stability-selection behavior.
- Seeded demonstration runs provide reproducible software-level evidence.
Limitations
What this work does not yet establish.
- Synthetic motor and flight-log inputs are used for the public demonstration.
- The software is not presented as flight-certified or physically validated.
- Real telemetry, weather, and higher-fidelity aerodynamic references remain future work.
Engineering decisions
Decisions that define the model boundary.
- Used typed Pydantic contracts between stages rather than passing unstructured dictionaries.
- Centralized rocket construction so geometry and common assumptions are not duplicated across simulation stages.
- Separated inexpensive recovery trade studies from the final full Monte Carlo confirmation analysis.
- Made the distinction between software verification and physical validation explicit throughout the project.
A note from the build
What stayed with me.
Hitting a target apogee in a demo is only one part of the picture. Stability, dispersion and recovery can change the decision, and synthetic results do not prove a vehicle is ready to fly.
I learned to make each stage hand off its assumptions clearly, and to separate software checks from the evidence needed to trust a real flight.
Next step
What I would investigate next.
Strengthen end-to-end integration tests, expand the motor × fin trade space, add real weather and telemetry inputs, cross-check transonic stability against higher-fidelity aerodynamic data, and expose the simulation as a foundation for guidance, control, and RL experiments.
Source and related work