Rocket SIM 6DOF

This is a Python Gym environment to simulate a 6 degrees of freedom (6DOF) rocket.

Features

• Full 6DOF rocket landing environment
• Realistic dynamics equations modeled on a rigid body assumption
• Interactive 3D visualization through PyVista
• Available actuators:
  • Thruster
  • Fins
• Wandb logging wrapper

Continuous Action Space

The environment employs a continuous action space, with the engine allowed to throttle between maxThrust and minThrust. The thrust is normalized to lie in the range [-1, +1] as best practice for algorithm convergence suggests. The engine is gimbaled by two angles, $\delta_y$ and $\delta_z$, around two hinge points, respectively moving the engine around the z and y axes.

Installation Instructions

1. Clone the repository:

   git clone https://github.com/yourusername/rocket-env.git
   cd rocket-env

2. Create and activate a virtual environment:

   python3 -m venv venv
   source venv/bin/activate

3. Install the package from the cloned repo:

   pip install .

Minimal Example with Rendering

Script to test the functionality of the 6DOF environment:

from rocket_env.envs import Rocket6DOF

# Initialize the environment
env = Rocket6DOF(render_mode="human")  # or "rgb_array"
obs, info = env.reset(seed=123)

while True:
    obs, rew, terminated, truncated, info, = env.step(env.action_space.sample())
    env.render()
    
    if terminated or truncated:
        env.reset()
        env.render()

env.close()

This will show the rendering: .

Minimal Example with Neural Control

import gymnasium as gym
from stable_baselines3 import PPO
from src.rocket_env.envs import Rocket6DOF

# Create the environment
# You can set render_mode="human" or "rgb_array" (if supported by your env)
env = Rocket6DOF(render_mode="human")

# Instantiate the agent
model = PPO("MlpPolicy", env, verbose=1)

# Train the agent
model.learn(total_timesteps=10000)

# Evaluate the agent
obs, info = env.reset(seed=123)
terminated = False
truncated = False

# Example evaluation loop
while not (terminated or truncated):
    action, _states = model.predict(obs, deterministic=True)
    
    # With Gymnasium, step returns (obs, reward, terminated, truncated, info)
    obs, reward, terminated, truncated, info = env.step(action)
    
    # Render if you want to see it
    env.render()
    
    if terminated or truncated:
        obs, info = env.reset()