Skip to main content

hydrogym.nek.AFC

The classical approaches for wing control Matching the structure from drl_repo_TEST/src/lib/AFC.py

AFC Objects​

class AFC()

A general object for classical AFC approaches

policy​

def policy(observation)

The control policy depends on the observation

predict​

def predict(observations, state, episode_start, deterministic)

Return the action to ENV based on the policy

OppoCtrl Objects​

class OppoCtrl(AFC)

Opposition control implementation

policy​

def policy(observation)

v = -alpha * (v - <v>)

BLCtrl Objects​

class BLCtrl(AFC)

Steady uniform blowing/suction

policy​

def policy(observation)

v = Psi

SinWave Objects​

class SinWave()

Imposing Sinusoidal wave to check the mean

load_node_info​

def load_node_info(Node_Info)

Get the node info to impose the Sinusoidal Wave

policy​

def policy(observation)

We define the V-Vel <==> 1

nameAgent​

@staticmethod
def nameAgent(nid, gllid, iface, ix, iy, iz)

Name the agent based on the GRID information

ZeroCtrl Objects​

class ZeroCtrl(AFC)

Zero action controller (no control)

make_afc_controller​

def make_afc_controller(env, ctrl_type="AFC")

Factory function to create an AFC controller compatible with integrate().

The controller adapts between array-based (NekEnv) and dict-based (NekParallelEnv, NekPettingZooEnv) formats.

To use an SB3 model, you can pass it directly to integrate(): from stable_baselines3 import PPO loaded_model = PPO.load("path/to/model") hgym.integrate(env, ..., controller=loaded_model)

Arguments:

  • env - Environment instance (NekEnv, NekParallelEnv, or NekPettingZooEnv)
  • ctrl_type - Algorithm name ("AFC", "OC", "BL", "SIN", "ZERO", or SB3 algorithm name)

Returns:

  • controller - Controller object with .predict() method, or None if not an AFC algorithm