2026_C_FD_SIM_Rocketpy

FD Rocketpy

Euroc Simulations

TODO:

Excel

TODO:

RocketPy_26

TODO:

Genetic Algorithm

Rocket Design Optimization Engine

Liquid and solid rocket design using algorithmic evolution. Reach a target apogee while minimizing drift and ensuring static stability. Built on RocketPy. Supports genetic, memetic, and Bayesian optimization with customizable wind layers.

Project Structure

GeneticAlgorithm/
├── archive
├── engine/
│   ├── algorithms/             # genetic.py, memetic.py, bayesian.py
│   ├── backtests/              # CSV logs from optimization runs
│   ├── data/                   # thrust curves, ullage CSVs, fluid properties
│   ├── environment_settings.py  # Environment factory and wind management
│   ├── evaluator.py            # Parallel evaluation and worker pool
│   ├── fitness.py              # Fitness function (apogee error, drift, stability)
│   ├── logger.py               # CSV logger
│   ├── main.py                 # Optimization entry point
│   ├── optimizer.py            # Abstract base class for algorithms
│   ├── printer.py              # Console output
│   ├── rocket_builder.py       # Rocket and motor assembly
│   ├── single_flight.py        # Single-design flight simulation
│   └── visualizer.py           # Post-run analysis and plotting
└── misc

Requirements

• Python 3.10+
• RocketPy >= 1.3.0
• NumPy, SciPy, scikit-learn, pandas, matplotlib, seaborn

pip install rocketpy numpy scipy scikit-learn pandas matplotlib seaborn

Usage

Running the Optimizer

python main.py [--algo {genetic,memetic,bayesian}] [options]

Key options:

Argument	Description	Default
--algo	Algorithm: genetic, memetic, bayesian	bayesian
--max_gen	Number of generations	15
--iter_per_gen	Population size per generation	20
--selection_method	Parent selection: rank, boltzmann, tournament	rank
--sp	Selection pressure (rank only)	1.5
--local_search_prob	Local search probability (memetic only)	0.35
--local_search_evals	Max local search steps (memetic only)	12
--random-wind	Use random wind layers (different each run unless seed is set)	off
--strong-wind	Apply constant 50 m/s wind for drift testing	off
--seed	Integer seed for reproducible random wind	None

Examples:

# Genetic algorithm, 10 generations, default population
python main.py --algo genetic --max_gen 10

# Bayesian optimization with 30 samples per generation
python main.py --algo bayesian --iter_per_gen 30

# Memetic algorithm with random wind layers
python main.py --algo memetic --random-wind

# Strong wind test (50 m/s from West)
python main.py --algo genetic --max_gen 5 --strong-wind

# Reproducible random wind using a seed
python main.py --algo bayesian --random-wind --seed 42

Testing a Single Design

Edit the DESIGN dictionary in single_flight.py to set your parameters, then:

python single_flight.py [--random-wind] [--strong-wind] [--upright] [--seed N]

Visualizing Results

Each run saves a CSV in backtests/. To plot:

# Auto-select latest log
python visualizer.py

# Specific file
python visualizer.py backtests/backtest_20260713181149_bayesian.csv

Generated plots: best fitness, mean fitness with error bars, best apogee, drift and stability margin evolution, parameter traces, correlation heatmap, fitness vs apogee scatter, feasible/infeasible counts.

How It Works

• Initialization: The algorithm creates an initial population of random designs within predefined bounds. Bayesian optimizer starts with a few random points before fitting its surrogate model.
• Evaluation: Each design is converted into a RocketPy rocket and simulated in parallel in a standard atmosphere with wind (if specified).
• Fitness: Rewards designs approaching the target apogee (3000 m) with low drift (bonus within 200 m), penalizing stability margins outside 1.5–5.0.
• Variation & Selection: Genetic/memetic algorithms apply crossover, mutation, and selection (memetic adds local random walk). Bayesian optimizer maximizes Expected Improvement via a Gaussian process.
• Elitism: The best individual is carried to the next generation.
• Logging: All evaluated designs are written to a timestamped CSV in backtests/.

Customization & Extension

• Add parameters: Extend VARIABLE_GENE_BOUNDS and INTEGER_GENES in main.py, then update RocketBuilder to use them.
• Change fitness: Modify fitness.py to adjust weights or add objectives.
• New wind models: Add cases in environment_settings.py's get_environment_parameters() function.
• New algorithm: Create a class in algorithms/ that inherits from OptimizerBase and implements run_generation(evaluator) to be automatically available via --algo.

License

This project is provided for educational and research purposes. Refer to the RocketPy license for its terms.

License

This project is provided for educational and research purposes. Refer to the RocketPy license for its terms.