Trajectory-Monte-Carlo

A compact, reproducible quasi-6-DoF projectile simulation framework built for large-scale Monte Carlo analysis and machine-learning dataset generation. The model couples simplified rigid-body dynamics with proportional-navigation steering and pitch-trim control.

Aerodynamic forces are computed from tabulated Cl/Cd values using 2-D interpolation, and pitching moments include static Cm(α), control-induced moment, and rate damping. Quaternion attitude propagation, fixed-step RK4 integration, and event-based ground impact detection ensure stable and consistent trajectory evolution.

This tool enables running thousands (or tens of thousands) of randomized trajectories (targets, winds, etc.) to evaluate miss-distance statistics or generate ML-ready datasets. Users may replace the placeholder/dummy aerodynamic tables with high-fidelity data from CFD or experiments.

The main simulation script (6DOF_Light_ML_Training.py) is an updated version emphasizing speed, reproducibility, and scalable batch execution using a hand-coded RK4 integrator. It can be extended to run 100,000+ trajectories for training neural-network surrogate models.

#EDIT: I have been playing around with the Aero Tables; current design space max trim AoA is ~14° from the given Cm,alpha table. If you edit the tables and want to see what happens at higher Angles of Attack, the design space is capped at 16°, you will need to update the aero tables accordingly. I also edited the numbers to have a more smooth representation. The Output gives great data for bulding a Neural Network.

Companion repo found here: https://github.com/Ultravis66/Guided-Trajectory-Neural-Net

Use Plotting.py to visualize the dataset and generate performance plots after running the simulation. Aerodynamic inputs in this repository are placeholder values; in practice these are typically replaced with CFD- or experimentally-derived tables.

Features

• Quasi-6-DoF rigid-body dynamics
  • 13-state model (position, velocity, quaternion attitude, angular rates)
• Guidance & control
  • Proportional-navigation-style lateral steering
  • Pitch-trim logic for altitude control
• Aerodynamics
  • Tabulated Cl/Cd lookups with 2-D interpolation
  • Cm(α) curve with control + damping contributions
  • Optional CP→CG moment arm
• Numerical integration
  • Hand-written fixed-step RK4
  • Stable quaternion normalization
• Monte Carlo engine
  • Randomized target locations & wind vectors
  • Batched execution for 1,000–100,000+ trajectories
  • Progress reporting, statistics, timing
• Data export
  • Summary CSV for machine learning
  • Individual trajectory .npz files (states + time histories)
• Plotting tools
  • Miss-distance histograms
  • CEP statistics
  • Target hit/miss maps
  • Trajectory visualizations

---

Author

if you use this work, please cite:

@misc{trajectoryMC2025,
author    = {Stolk, Mitchell},
title     = {Trajectory-Monte-Carlo: Quasi-6DoF Projectile Simulator and ML Dataset Generator},
year      = {2025},
publisher = {GitHub},
url       = {https://github.com/ultravis66/Trajectory-Monte-Carlo}
}