# -*- coding: utf-8 -*-
# Copyright 2021 IRT Saint Exupéry, https://www.irt-saintexupery.com
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License version 3 as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public License
# along with this program; if not, write to the Free Software Foundation,
# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
# Contributors:
# INITIAL AUTHORS - API and implementation and/or documentation
# :author: Francois Gallard
# OTHER AUTHORS - MACROSCOPIC CHANGES
"""A Bi-level formulation."""
from __future__ import division, unicode_literals
import logging
from typing import Any, Dict, Iterable, List, Optional, Sequence, Tuple, Union
from gemseo.algos.design_space import DesignSpace
from gemseo.core.chain import MDOChain, MDOParallelChain
from gemseo.core.coupling_structure import MDOCouplingStructure
from gemseo.core.discipline import MDODiscipline
from gemseo.core.execution_sequence import ExecutionSequence
from gemseo.core.formulation import MDOFormulation
from gemseo.core.function import MDOFunction
from gemseo.core.json_grammar import JSONGrammar
from gemseo.core.mdo_scenario import MDOScenarioAdapter
from gemseo.core.scenario import Scenario
from gemseo.mda.mda import MDA
from gemseo.mda.mda_factory import MDAFactory
LOGGER = logging.getLogger(__name__)
[docs]class BiLevel(MDOFormulation):
"""A bi-level formulation.
This formulation draws an optimization architecture
that involves multiple optimization problems to be solved
to obtain the solution of the MDO problem.
Here,
at each iteration on the global design variables,
the bi-level MDO formulation implementation performs:
1. a first MDA to compute the coupling variables,
2. several disciplinary optimizations on the local design variables in parallel,
3. a second MDA to update the coupling variables.
"""
SYSTEM_LEVEL = "system"
SUBSCENARIOS_LEVEL = "sub-scenarios"
LEVELS = (SYSTEM_LEVEL, SUBSCENARIOS_LEVEL)
def __init__(
self,
disciplines, # type: Sequence[MDODiscipline]
objective_name, # type: str
design_space, # type: DesignSpace
maximize_objective=False, # type: bool
mda_name="MDAChain", # type: str
parallel_scenarios=False, # type: bool
multithread_scenarios=True, # type: bool
apply_cstr_tosub_scenarios=True, # type: bool
apply_cstr_to_system=True, # type: bool
reset_x0_before_opt=False, # type: bool
**mda_options # type: Any
): # type: (...) -> None
"""
Args:
mda_name: The name of the MDA class to be used.
parallel_scenarios: If True, the sub-scenarios are run in parallel.
multithread_scenarios: If True and parallel_scenarios=True,
the sub-scenarios are run in parallel using multi-threading;
if False and parallel_scenarios=True, multi-processing is used.
apply_cstr_tosub_scenarios: If True,
the :meth:`.add_constraint` method adds the constraint
to the optimization problem of the sub-scenario
capable of computing the constraint.
apply_cstr_to_system: If True,
the :meth:`.add_constraint` method adds the constraint
to the optimization problem of the system scenario.
reset_x0_before_opt: If True,
restart the sub optimizations from the initial guesses,
otherwise warm start them.
**mda_options: The options passed to the MDA at construction.
"""
super(BiLevel, self).__init__(
disciplines,
objective_name,
design_space,
maximize_objective=maximize_objective,
)
self._shared_dv = list(design_space.variables_names)
self.mda1 = None
self.mda2 = None
self.reset_x0_before_opt = reset_x0_before_opt
self.scenario_adapters = []
self.chain = None
self._mda_factory = MDAFactory()
self._apply_cstr_to_system = apply_cstr_to_system
self._apply_cstr_tosub_scenarios = apply_cstr_tosub_scenarios
self.__parallel_scenarios = parallel_scenarios
self._multithread_scenarios = multithread_scenarios
self.couplstr = MDOCouplingStructure(self.get_sub_disciplines())
# Create MDA
self._build_mdas(mda_name, **mda_options)
# Create MDOChain : MDA1 -> sub scenarios -> MDA2
self._build_chain()
# Cleanup design space
self._update_design_space()
# Builds the objective function on top of the chain
self._build_objective_from_disc(self._objective_name)
def _build_scenario_adapters(
self,
output_functions=False, # type: bool
pass_nonshared_var=False, # type: bool
adapter_class=MDOScenarioAdapter, # type:MDOScenarioAdapter
**adapter_options
): # type: (...) -> List[MDOScenarioAdapter]
"""Build the MDOScenarioAdapter required for each sub scenario.
This is used to build the self.chain.
Args:
output_functions: If True,
then the optimization functions are outputs of the adapter
pass_nonshared_var: If True,
the non-shared design variables are inputs of the scenarios adapters.
adapter_class: The class of the adapters.
**adapter_options: The options for the adapters initialization.
Returns:
The adapters for the sub-scenarios.
"""
adapters = []
# coupled sub-disciplines
couplings = self.couplstr.strong_couplings()
mda2_inpts = self._get_mda2_inputs()
shared_dv = set(self._shared_dv)
for scenario in self.get_sub_scenarios():
# Get the I/O names of the sub-scenario top-level disciplines
top_disc = scenario.formulation.get_top_level_disc()
top_inputs = [
inpt for disc in top_disc for inpt in disc.get_input_data_names()
]
top_outputs = [
outpt for disc in top_disc for outpt in disc.get_output_data_names()
]
mda1_outputs = self.mda1.get_output_data_names() if self.mda1 else []
# All couplings of the scenarios are taken from the MDA
sc_allins = list(
# Add shared variables from system scenario driver
set(top_inputs)
& set(set(couplings) | shared_dv | set(mda1_outputs))
)
if pass_nonshared_var:
nonshared_var = scenario.design_space.variables_names
sc_allins = list(set(sc_allins) | set(top_inputs) & set(nonshared_var))
# Output couplings of scenario are given to MDA for speedup
if output_functions:
opt_problem = scenario.formulation.opt_problem
sc_outvars = opt_problem.objective.outvars
sc_constraints = opt_problem.get_constraints_names()
sc_out_coupl = sc_outvars + sc_constraints
else:
sc_out_coupl = list(set(top_outputs) & set(couplings + mda2_inpts))
# Add private variables from disciplinary scenario design space
sc_allouts = sc_out_coupl + scenario.design_space.variables_names
adapter = adapter_class(scenario, sc_allins, sc_allouts, **adapter_options)
adapters.append(adapter)
return adapters
@staticmethod
def _get_mda2_inputs(): # type: (...) -> List[str]
"""Return the inputs of the second MDA.
Returns:
The inputs of the second MDA.
"""
return []
[docs] @classmethod
def get_sub_options_grammar(
cls, **options # type: str
): # type: (...) -> JSONGrammar
main_mda = options.get("mda_name")
if main_mda is None:
raise ValueError(
"'mda_name' option is required to deduce the sub options of BiLevel !"
)
factory = MDAFactory().factory
return factory.get_options_grammar(main_mda)
[docs] @classmethod
def get_default_sub_options_values(
cls, **options # type: str
): # type: (...) -> Optional[Dict[str,Optional[Union[str,int,float,bool]]]]
main_mda = options.get("mda_name")
if main_mda is None:
raise ValueError(
"'mda_name' option is required to deduce the sub options of BiLevel !"
)
factory = MDAFactory().factory
return factory.get_default_options_values(main_mda)
def _build_mdas(
self,
mda_name, # type: str
**mda_options # type:Optional[Union[str,int,float,bool]]
): # type: (...) -> None
"""Build the chain on top of which all functions are built.
This chain is: MDA -> MDOScenarios -> MDA.
Args:
mda_name: The class name of the MDA.
**mda_options: The options passed to the MDA.
"""
disc_mda1 = self.couplstr.strongly_coupled_disciplines()
if len(disc_mda1) > 0:
self.mda1 = self._mda_factory.create(mda_name, disc_mda1, **mda_options)
self.mda1.warm_start = True
else:
LOGGER.warning(
"No strongly coupled disciplines detected, "
" MDA1 is deactivated in the BiLevel formulation"
)
disc_mda2 = self.get_sub_disciplines()
self.mda2 = self._mda_factory.create(mda_name, disc_mda2, **mda_options)
self.mda2.warm_start = False
def _build_chain_dis_sub_opts(
self,
): # type: (...) -> Tuple[Union[List,MDA], List[MDOScenarioAdapter]]
"""Initialize the chain of disciplines and the sub-scenarios.
Returns:
The first MDA (if exists) and the sub-scenarios.
"""
chain_dis = []
if self.mda1 is not None:
chain_dis = [self.mda1]
sub_opts = self.scenario_adapters
return chain_dis, sub_opts
def _build_chain(self): # type: (...) -> None
"""Build the chain on top of which all functions are built.
This chain is: MDA -> MDOScenarios -> MDA.
"""
# Build the scenario adapters to be chained with MDAs
adapter_opt = {"reset_x0_before_opt": self.reset_x0_before_opt}
self.scenario_adapters = self._build_scenario_adapters(**adapter_opt)
chain_dis, sub_opts = self._build_chain_dis_sub_opts()
if self.__parallel_scenarios:
use_threading = self._multithread_scenarios
par_chain = MDOParallelChain(sub_opts, use_threading=use_threading)
chain_dis += [par_chain, self.mda2]
else:
# Chain MDA -> scenarios exec -> MDA
chain_dis += sub_opts + [self.mda2]
self.chain = MDOChain(chain_dis, name="bilevel_chain")
if not self.reset_x0_before_opt and self.mda1 is not None:
run_mda1_orig = self.mda1._run
def _run_mda():
"""Redefine mda1 execution to warm start the chain with previous x_local
opt."""
# TODO : Define a pre run method to be overloaded in MDA maybe
# Or use observers at the system driver level to pass the local
# vars
for scenario in self.get_sub_scenarios():
x_loc_d = scenario.design_space.get_current_x_dict()
for indata, x_loc in x_loc_d.items():
if self.mda1.is_input_existing(indata):
if x_loc is not None:
self.mda1.local_data[indata] = x_loc
return run_mda1_orig()
self.mda1._run = _run_mda
def _update_design_space(self): # type: (...) -> None
"""Update the design space by removing the coupling variables."""
self._set_defaultinputs_from_ds()
self._remove_sub_scenario_dv_from_ds()
self._remove_couplings_from_ds()
self._remove_unused_variables()
def _remove_couplings_from_ds(self): # type: (...) -> None
"""Removes the coupling variables from the design space."""
if hasattr(self.mda2, "strong_couplings"):
# Otherwise, the MDA2 may be a user provided MDA
# Which manages the couplings internally
couplings = self.mda2.strong_couplings
design_space = self.opt_problem.design_space
for coupling in couplings:
if coupling in design_space.variables_names:
design_space.remove_variable(coupling)
[docs] def get_top_level_disc(self): # type: (...) -> List[MDODiscipline]
return [self.chain]
[docs] def get_expected_workflow(
self,
): # type: (...) -> List[ExecutionSequence,Tuple[ExecutionSequence]]
return self.chain.get_expected_workflow()
[docs] def get_expected_dataflow(
self,
): # type: (...) -> List[Tuple[MDODiscipline,MDODiscipline,List[str]]]
return self.chain.get_expected_dataflow()
[docs] def add_constraint(
self,
output_name, # type: str
constraint_type=MDOFunction.TYPE_EQ, # type: str
constraint_name=None, # type: Optional[str]
value=None, # type: Optional[float]
positive=False, # type: bool
levels=None, # type: Optional[List[str]]
): # type: (...) -> None
"""Add a constraint to the formulation.
Args:
levels: The levels at which the constraint is to be added
(sublist of Bilevel.LEVELS).
By default the policy set at the initialization
of the formulation is enforced.
"""
# If the constraint levels are not specified the initial policy is enforced.
if levels is None:
if self._apply_cstr_to_system:
self._add_system_level_constraint(
output_name, constraint_type, constraint_name, value, positive
)
if self._apply_cstr_tosub_scenarios:
self._add_sub_level_constraint(
output_name, constraint_type, constraint_name, value, positive
)
# Otherwise the constraint is applied at the specified levels.
elif not isinstance(levels, list) or not set(levels) <= set(BiLevel.LEVELS):
raise ValueError(
"Constraint levels must be a sublist of {}".format(BiLevel.LEVELS)
)
elif not levels:
LOGGER.warning("Empty list of constraint levels, constraint not added")
else:
if BiLevel.SYSTEM_LEVEL in levels:
self._add_system_level_constraint(
output_name, constraint_type, constraint_name, value, positive
)
if BiLevel.SUBSCENARIOS_LEVEL in levels:
self._add_sub_level_constraint(
output_name, constraint_type, constraint_name, value, positive
)
def _add_system_level_constraint(
self,
output_name, # type: str
constraint_type=MDOFunction.TYPE_EQ, # type: str
constraint_name=None, # type: Optional[str]
value=None, # type: Optional[float]
positive=False, # type: bool
): # type: (...) -> None
"""Add a constraint at the system level.
Args:
output_name: The name of the output to be used as a constraint.
For instance, if g_1 is given and constraint_type="eq",
g_1=0 will be added as a constraint to the optimizer.
constraint_type: The type of constraint,
either "eq" for equality constraint or "ineq" for inequality constraint.
constraint_name: The name of the constraint to be stored,
If None, the name is generated from the output name.
value: The value of activation of the constraint.
If None, the value is equal to 0.
positive: If True, the inequality constraint is positive.
"""
super(BiLevel, self).add_constraint(
output_name, constraint_type, constraint_name, value, positive
)
def _add_sub_level_constraint(
self,
output_name, # type: str
constraint_type=MDOFunction.TYPE_EQ, # type: str
constraint_name=None, # type: Optional[str]
value=None, # type: Optional[float]
positive=False, # type: bool
): # type: (...) -> None
"""Add a constraint at the sub-scenarios level.
Args:
output_name: The name of the output to be used as a constraint.
For instance, if g_1 is given and constraint_type="eq",
g_1=0 will be added as a constraint to the optimizer.
constraint_type: The type of constraint,
either "eq" for equality constraint or "ineq" for inequality constraint.
constraint_name: The name of the constraint to be stored,
If None, the name is generated from the output name.
value: The value of activation of the constraint.
If None, the value is equal to 0.
positive: If True, the inequality constraint is positive.
"""
added = False
outputs_list = self._check_add_cstr_input(output_name, constraint_type)
for scen in self.get_sub_scenarios():
if self._scenario_computes_outputs(scen, outputs_list):
scen.add_constraint(
outputs_list, constraint_type, constraint_name, value, positive
)
added = True
if not added:
raise ValueError(
"No sub scenario has an output named {} "
"cannot create such a constraint.".format(output_name)
)
@staticmethod
def _scenario_computes_outputs(
scenario, # type: Scenario
output_names, # type: Iterable[str]
): # type: (...) -> bool
"""Check if the top level disciplines compute the given outputs.
Args:
output_names: The names of the variable to check.
scenario: The scenario to be tested.
Return:
True if the top level disciplines compute the given outputs.
"""
for disc in scenario.formulation.get_top_level_disc():
if disc.is_all_outputs_existing(output_names):
return True
return False
