Skip to main content

Pinball Flow Environment

This module provides pinball (multi-cylinder) flow CFD environments for reinforcement learning.

PinballBase Objects​

class PinballBase(MaiaFlowEnv)

Base class for pinball flow environments with Hugging Face integration.

The pinball configuration consists of multiple cylinders arranged in a triangular pattern. Forces from all cylinders are summed for the reward.

The reward is computed as: reward = -|sum(C_D)| - omega * |sum(C_L)|

where C_D and C_L are summed across all cylinders.

__init__​

def __init__(env_config: Dict)

Initialize the pinball base environment.

Arguments:

  • env_config - Environment configuration dictionary.

get_reward​

def get_reward() -> Tuple[float, Dict]

Compute the reward based on total aerodynamic force coefficients.

Aggregates forces from all cylinders in the pinball configuration.

Returns:

Tuple containing:

  • reward: Scalar reward based on total drag and lift
  • obj_dict: Dictionary with force coefficient information

Pinball Objects​

class Pinball(PinballBase)

Rotary pinball environment with rotational actuation.

This environment controls flow using cylinder rotation. Each cylinder can rotate independently.

__init__​

def __init__(env_config: Dict)

Initialize the pinball environment.

Arguments:

  • env_config - Environment configuration dictionary.

convert_action​

def convert_action(action: np.ndarray) -> np.ndarray

Convert RL action to CFD actuation format.

Arguments:

  • action - Action array from the RL agent.

Returns:

Action sequence for the CFD solver.

JetPinball Objects​

class JetPinball(PinballBase)

Jet-actuated pinball environment.

This environment controls flow using synthetic jets on the cylinders. Jets are paired for zero net mass flux actuation.

Attributes:

  • numJetsInSimulation - Number of jet actuators in the CFD simulation.

__init__​

def __init__(env_config: Dict)

Initialize the jet pinball environment.

Arguments:

  • env_config - Environment configuration dictionary.

convert_action​

def convert_action(action: np.ndarray) -> List[float]

Convert RL action to CFD actuation sequence.

Implements zero net mass flux by pairing jets with opposite signs.

Arguments:

  • action - Action array from the RL agent.

Returns:

Actuation sequence for the CFD solver.