### smarties coupled with LESGO

Reinforcement learning wall-model (RLWM) for large eddy simulation smarties library coupled with LESGO

-> You should first download and install smarties from https://github.com/cselab/smarties

-> The /apps folders contains the routines to run the model, for example LESGO with coupling routines (smarties_stat.f90 and app_main.f90). In particular, the functions send_recv_state_action is the main important since it enables to communicate with agents and get the action. CAREFUL : use runfile_training with care since it erases the existing weights and biases in the /runs folder if it exists. Use runfile_exec to use the RLWM without retraining.

Details can be found in :

Vadrot, A., Yang, X. I., Bae, H. J., & Abkar, M. (2023). Log-law recovery through reinforcement-learning wall model for large eddy simulation. Physics of Fluids, 35(5).

-> You can execute the code in two ways: either train your own model or run the existing model. In case you train your own model, it will create a new folder in /runs and print the output of LESGO for all the simulations that are used for training simulation_000_.... When testing it will use the weights and biases of RL network (agent_00_net***) contained in the corresponding /runs folder. CAREFUL: use runfile_training with care since it erases the existing weights and biases in the /runs folder if they exist.

-> To cite this repository, reference the paper:

@article{vadrot2023log,

title={Log-law recovery through reinforcement-learning wall model for large eddy simulation}, author={Vadrot, Aur{'e}lien and Yang, Xiang IA and Bae, H Jane and Abkar, Mahdi}, journal={Physics of Fluids}, volume={35}, number={5}, year={2023}, publisher={AIP Publishing}

}