LocalSubSystemALC¶

Source: Distributed_Design_Optimizer/subsystem/LocalSubSystemALC.py

Augmented Lagrangian Coordination local subsystem module.

This module provides the local subsystem implementation for the standard Augmented Lagrangian Coordination method.

Classes¶

LocalSubSystemALC¶

Inherits from: LocalSubSystemBasis

Local subsystem implementation using Augmented Lagrangian Coordination.

Implements the Augmented Lagrangian Coordination (ALC) method for distributed optimization. Manages coupling parameters, coordination multipliers, weights, and convergence tracking for iterative decomposition-based optimization.

Methods¶

init(self, id: str, level: int, neighborid: List[str], analysis: AnalysisInterface, localobjective: LocalObjectiveInterface, localconstraints: LocalConstraintsInterface, optimization: OptimizationInterface, local_convergenceindicator_innerloop: Local_ConvergenceIndicator_Innerloop_Interface, local_convergenceindicator_outerloop: Local_ConvergenceIndicator_Outerloop_Interface, updatecouplingparametermethod_outerloop: UpdateCouplingParameterMethodInterface) → None

📐 Pseudocode: Require hyperparameters γ, β and initial coupling parameters (Lines 1, 2, 3) → Augmented Lagrangian Coordination

Create a new LocalSubSystemALC instance.

Args:

id: Identifier for the subsystem.
level: Level identifier for the subsystem in the hierarchy.
neighborid: List of identifiers for neighboring subsystems.
analysis: Analysis interface for subsystem evaluation.
localobjective: Local objective function class.
localconstraints: Local constraint functions class.
optimization: Optimization interface for solving local problems.
local_convergenceindicator_innerloop: Local convergence indicator for inner loop.
local_convergenceindicator_outerloop: Local convergence indicator for outer loop.
updatecouplingparametermethod_outerloop: Method for updating coupling parameters in outer loop.

validate_inputs(self) → None

Validate the components handed to this subsystem by the ALC coordination method.

Validates that the outer-loop update coupling parameter method and the local inner/outer loop convergence indicators are compatible with ALC.

Raises:

ValueError: If a provided component is not compatible with ALC.

append_Controller(self) → None

Update local subsystems by appending controller.

mapToController(self) → None

Pass, since no controller in ALC.

evaluateCoordinationObjective(self) → None

📐 Pseudocode: Solve ALC augmented Lagrangian optimization (Line 10) → Augmented Lagrangian Coordination

Evaluate the coordination objective.

evaluate_Gradient_CoordinationObjective(self) → None

Evaluate the analytical gradient of the coordination objective.

Assumes the mapped responses Jacobian was evaluated already.

evaluateCoordinationEqualityConstraint(self) → None

Evaluate the coordination equality constraint.

evaluateCoordinationInequalityConstraint(self) → None

Evaluate the coordination inequality constraint.

evaluate_Jacobian_CoordinationEqualityConstraints(self) → None

Evaluate the Jacobian of the coordination equality constraints.

ALC does not have coordination equality constraints, hence no-op.

evaluate_Jacobian_CoordinationInEqualityConstraints(self) → None

Evaluate the Jacobian of the coordination inequality constraints.

ALC does not have coordination inequality constraints, hence no-op.

prepare_OptimizationProblem(self) → None

Prepare the optimization problem.

postprocess_Optimization(self) → None

Postprocess the optimization.

return_initialized_CouplingParameters(self) → List[CouplingParametersALC]

Return initialized coupling parameters.

Returns:

List of initialized CouplingParametersALC, one per neighbor.

initializeCouplingParameters_before_CopyToMiddleLevel(self) → None

Initialize coupling parameters before copying from neighboring subsystem.

Create empty placeholders for coupling parameters, i.e. communicated quantities in the algorithms, Lagrange multipliers, penalty weights, ... The sizes are read by already initialized quantities from the inner loop iteration.

initializeCouplingParameters_after_CopyFromMiddleLevel(self) → None

📐 Pseudocode: Require hyperparameters γ, β and initial coupling parameters (Lines 1, 2, 3) → Augmented Lagrangian Coordination

Initialize coupling parameters after copying from neighboring subsystems.

Create empty placeholders for coupling parameters, i.e. communicated quantities in the algorithms, Lagrange multipliers, penalty weights, ... The sizes are read by already initialized quantities from the inner loop iteration.

initializeCouplingParameters_after_Second_CopyFromMiddleLevel(self) → None

Initialize coupling parameters after two communication rounds between subsystems.

prepare_updateCouplingParameters(self) → None

Prepare coupling parameters before update operations.

updateCouplingParameters_innerLoop(self) → None

Update coupling parameters in the inner loop.

updateCouplingParameters_outerLoop(self) → None

📐 Pseudocode: Dual Update (Lagrange multiplier update) (Line 20), Penalty Update (weight update) (Line 20) → Augmented Lagrangian Coordination

Update coupling parameters during outer loop iteration.

evaluate_Inconsistencies(self) → None

📐 Pseudocode: Outerloop convergence and return (Lines 22, 23, 24, 25) → Augmented Lagrangian Coordination

Compute the difference between stored coupling and mapped variables.

Compares to the latest available data from a subsystem. Delegates to the base class implementation which computes inconsistency vectors for the mapped-response side, coupling-variable side, shared design variables, and target shared design variables of each coupling circle.

return_initialized_Inconsistencies(self) → List[InConsistencySize]

Return initialized inconsistencies.

Returns:

List of initialized InConsistencySize, one per coupling parameter.

set_CoordinationMultipliers_MappedResponse_Minus_CopyCouplingVariable(self, neighborid: str, multipliersin: List[float]) → None

📐 Pseudocode: Dual Update (Lagrange multiplier update) (Line 20) → Augmented Lagrangian Coordination

Set coordination multipliers for the mapped response minus copied coupling variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
multipliersin: List of multiplier values to set.

set_CoordinationMultipliers_CopyMappedResponse_Minus_CouplingVariable(self, neighborid: str, multipliersin: List[float]) → None

📐 Pseudocode: Dual Update (Lagrange multiplier update) (Line 20) → Augmented Lagrangian Coordination

Set coordination multipliers for the copied mapped response minus coupling variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
multipliersin: List of multiplier values to set.

set_CoordinationMultipliers_SharedDesignVariable_Minus_CopyTargetSharedDesignVariable(self, neighborid: str, multipliersin: List[float]) → None

📐 Pseudocode: Dual Update (Lagrange multiplier update) (Line 20) → Augmented Lagrangian Coordination

Set coordination multipliers for the shared design variable minus copied target shared design variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
multipliersin: List of multiplier values to set.

set_CoordinationMultipliers_CopySharedDesignVariable_Minus_TargetSharedDesignVariable(self, neighborid: str, multipliersin: List[float]) → None

📐 Pseudocode: Dual Update (Lagrange multiplier update) (Line 20) → Augmented Lagrangian Coordination

Set coordination multipliers for the copied shared design variable minus target shared design variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
multipliersin: List of multiplier values to set.

set_CoordinationWeights_MappedResponse_Minus_CopyCouplingVariable(self, neighborid: str, penaltyweightsin: List[float]) → None

📐 Pseudocode: Penalty Update (weight update) (Line 20) → Augmented Lagrangian Coordination

Set coordination weights for the mapped response minus copied coupling variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
penaltyweightsin: List of penalty weight values to set.

set_CoordinationWeights_CopyMappedResponse_Minus_CouplingVariable(self, neighborid: str, penaltyweightsin: List[float]) → None

📐 Pseudocode: Penalty Update (weight update) (Line 20) → Augmented Lagrangian Coordination

Set coordination weights for the copied mapped response minus coupling variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
penaltyweightsin: List of penalty weight values to set.

set_CoordinationWeights_SharedDesignVariable_Minus_CopyTargetSharedDesignVariable(self, neighborid: str, penaltyweightsin: List[float]) → None

📐 Pseudocode: Penalty Update (weight update) (Line 20) → Augmented Lagrangian Coordination

Set coordination weights for the shared design variable minus copied target shared design variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
penaltyweightsin: List of penalty weight values to set.

set_CoordinationWeights_CopySharedDesignVariable_Minus_TargetSharedDesignVariable(self, neighborid: str, penaltyweightsin: List[float]) → None

📐 Pseudocode: Penalty Update (weight update) (Line 20) → Augmented Lagrangian Coordination

Set coordination weights for the copied shared design variable minus target shared design variable inconsistency of a neighbor.

Args:

neighborid: Identifier of the neighboring subsystem.
penaltyweightsin: List of penalty weight values to set.

update_state(self, other_subsystem: LocalSubSystemALC) → None

Update the state of this LocalSubSystemALC instance with values from another instance.

This method is necessary for multiprocessing. When subsystems are executed in parallel using multiprocessing.Pool, they are serialized and deserialized, creating new objects in separate memory spaces. After parallel execution completes, this method updates the original object's attribute values while preserving their memory addresses.

The update preserves memory addresses by modifying attribute contents in-place where possible, rather than reassigning references. This is essential for maintaining object identity across the multiprocessing boundary.

Args:

other_subsystem: The source LocalSubSystemALC containing
updated values from parallel execution.