![Trajectory](https://private-user-images.githubusercontent.com/3532595/390734362-424d0c10-5d9c-4f9d-8dc7-4d51a436cd43.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.jtACdevvRSzg5WgfIn1XUN3lK_-FN6Uc0E5gnAZOE1w)
Repository for the development of my master thesis on control of launch vehicles descent and landing through reinforcement learning actors.
Separate Gym environment available.
In this repository a full 6DOF rocket landing environment is developed, with realistic dynamics equation modeled on a rigid body assumption.
The environment employes a continuous action space, with the engine allowed to throttle between maxThrust
and minThrust
. The thrust was normalized to lie in the range [-1, +1]
as best practice for convergence of the algorithms suggest. The engine is gimbaled by two angles
To run the algorithm in a Docker container follow these steps:
-
Clone the repository
-
Build the docker image
docker build -t rl_rocket_docker .
-
Get your Wandb API key from wandb.ai/authorize
-
Start the docker container passing the API key as an environmental variable (paste it in place of
$YOUR_API_KEY$
)docker run -e WANDB_API_KEY=$YOUR_API_KEY$ -it rl_rocket_docker