Genetic Algorithm for Biomechanical Movement Optimization

A Python-based optimization framework using Genetic Algorithms and CMA-ES (Covariance Matrix Adaptation Evolution Strategy) to optimize biomechanical movements in OpenSim models. This project simulates and optimizes muscle activation patterns for various movements including jumping, counter-movement jumps, and flips. This project is developed only for learning purposes and not suitable for research.

Features

• Multiple Optimization Algorithms

  • Genetic Algorithm with customizable mutation, crossover, and selection
  • CMA-ES (Covariance Matrix Adaptation Evolution Strategy) solver

• Biomechanical Movement Types

  • Jumper: Basic vertical jump optimization
  • cmjJumper: Counter-movement jump optimization
  • Flipper: Backflip/flip movement optimization

• Parallel Processing

  • Multi-worker support for faster simulations
  • Automatic CPU core detection or manual worker configuration

• Flexible Muscle Control

  • Muscle group-based control system
  • Time-based activation patterns
  • Support for 8 major muscle groups:
    • Quadriceps
    • Hamstrings
    • Glutes
    • Calves
    • Tibialis Anterior
    • Trunk Flexors
    • Trunk Extensors
    • Hip Flexors

Requirements

This project requires:

• Python 3.x
• OpenSim (opensim-python)
• NumPy
• CMA (for CMA-ES solver)
• Multiprocessing support

See requirements.txt for the complete list of dependencies.

Installation

1. Clone the repository:

git clone https://github.com/Talhauzumcu/GeneticAlgorithm.git
cd GeneticAlgorithm

2. Install Miniconda (if not already installed):

   • Download from Miniconda official site
   • Follow the installation instructions for your operating system

3. Create and activate a conda environment:

conda create -n opensim-env python=3.x
conda activate opensim-env

4. Install OpenSim:

conda install -c opensim-org opensim

5. Install the remaining dependencies:

pip install -r requirements.txt

Usage

Basic Usage

Run the main script to start optimization:

python main.py

Configuration

Edit the parameters in main.py to customize the optimization:

POPULATION_SIZE = 200          # Number of individuals in each generation
GENERATION_COUNT = 1000        # Number of generations to evolve
MUTATION_RATE = 0.02          # Probability of mutation (0-1)
INTEGRATION_DURATION = 2       # Simulation time in seconds
OVERLAP = 5                    # Elite individuals to carry forward
RANDOM = 0                     # Random individuals to introduce per generation
N_WORKERS = 8                  # Number of parallel workers

Using Different Movement Types

Choose your movement type in main.py:

from objects import Jumper, cmjJumper, Flipper

# Select one:
pop_object = Jumper      # For basic jump
pop_object = cmjJumper   # For counter-movement jump
pop_object = Flipper     # For flip movements

Using Genetic Algorithm

from geneticSolver import geneticSolver

genetic_solver = geneticSolver(
    model_path='./models/your_model.osim',
    population_object=Jumper,
    initial_population=None,  # or provide custom population
    population_size=200,
    generation_count=1000,
    mutation_rate=0.02,
    overlap=5,
    random=0,
    initial_time=0.0,
    final_time=2.0,
    n_workers=8,
    termination_function=your_termination_function
)

Using CMA-ES Solver

from cmaSolver import cmaSolver

cma_solver = cmaSolver(
    model_path='./models/your_model.osim',
    population_object=Jumper,
    population_size=200,
    generation_count=1000,
    initial_time=0.0,
    final_time=2.0,
    n_workers=8,
    termination_function=your_termination_function,
    sigma0=0.3  # Initial standard deviation
)

Output

Results are saved as JSON files containing:

• Generation number
• Individual ID
• Fitness score
• Muscle activation patterns (genom)

Example filename: flipper_gen100_15_1.84.json

• gen100: Generation 100
• 15: Individual ID 15
• 1.84: Fitness score of 1.84

Custom Termination Functions

Define custom termination conditions for your simulations:

def termination_function(model, state):
    pelvis_ty = model.getCoordinateSet().get('pelvis_ty').getValue(state)
    return pelvis_ty < 0.25  # Stop if pelvis height below 0.25

Authors

• Talha Uzumcu