public static void main(String[] args) {
// see an example with detailed explanations -- package
// opdytsintegration.example.networkparameters.RunNetworkParameters
Config config = ConfigUtils.loadConfig(EQUIL_DIR + "/config.xml");
config.controler().setOutputDirectory(OUT_DIR);
config.controler().setOverwriteFileSetting(OverwriteFileSetting.deleteDirectoryIfExists);
// config.plans().setInputFile("relaxed_plans.xml.gz");
config.plans().setInputFile("plans2000.xml.gz");
// == default config has limited inputs
StrategyConfigGroup strategies = config.strategy();
strategies.clearStrategySettings();
config.changeMode().setModes(new String[] {"car", "pt"});
StrategySettings modeChoice = new StrategySettings();
modeChoice.setStrategyName(
DefaultPlanStrategiesModule.DefaultStrategy.ChangeSingleTripMode.name());
modeChoice.setWeight(0.1);
config.strategy().addStrategySettings(modeChoice);
StrategySettings expChangeBeta = new StrategySettings();
expChangeBeta.setStrategyName(DefaultPlanStrategiesModule.DefaultSelector.ChangeExpBeta.name());
expChangeBeta.setWeight(0.9);
config.strategy().addStrategySettings(expChangeBeta);
for (PlanCalcScoreConfigGroup.ActivityParams params :
config.planCalcScore().getActivityParams()) {
params.setTypicalDurationScoreComputation(
PlanCalcScoreConfigGroup.TypicalDurationScoreComputation.relative);
}
// config.qsim().setTrafficDynamics( QSimConfigGroup.TrafficDynamics.withHoles );
//
// if ( config.qsim().getTrafficDynamics()== QSimConfigGroup.TrafficDynamics.withHoles ) {
// config.qsim().setInflowConstraint(QSimConfigGroup.InflowConstraint.maxflowFromFdiag);
// }
config.qsim().setUsingFastCapacityUpdate(true);
// ==
Scenario scenario = KNBerlinControler.prepareScenario(true, false, config);
double time = 6 * 3600.;
for (Person person : scenario.getPopulation().getPersons().values()) {
Plan plan = person.getSelectedPlan();
Activity activity = (Activity) plan.getPlanElements().get(0);
activity.setEndTime(time);
time++;
}
// ==
// this is something like time bin generator
int startTime = 0;
int binSize = 3600; // can this be scenario simulation end time.
int binCount = 24; // to me, binCount and binSize must be related
TimeDiscretization timeDiscretization = new TimeDiscretization(startTime, binSize, binCount);
Set<String> modes2consider = new HashSet<>();
modes2consider.add("car");
modes2consider.add("bike");
OpdytsModalStatsControlerListener stasControlerListner =
new OpdytsModalStatsControlerListener(modes2consider, EQUIL);
// following is the entry point to start a matsim controler together with opdyts
MATSimSimulator<ModeChoiceDecisionVariable> simulator =
new MATSimSimulator<>(new MATSimStateFactoryImpl<>(), scenario, timeDiscretization);
simulator.addOverridingModule(
new AbstractModule() {
@Override
public void install() {
// add here whatever should be attached to matsim controler
// some stats
addControlerListenerBinding().toInstance(stasControlerListner);
// from KN
addControlerListenerBinding().to(KaiAnalysisListener.class);
bind(CharyparNagelScoringParametersForPerson.class)
.to(EveryIterationScoringParameters.class);
}
});
// this is the objective Function which returns the value for given SimulatorState
// in my case, this will be the distance based modal split
ObjectiveFunction objectiveFunction =
new ModeChoiceObjectiveFunction(
EQUIL); // in this, the method argument (SimulatorStat) is not used.
// search algorithm
int maxIterations =
10; // this many times simulator.run(...) and thus controler.run() will be called.
int maxTransitions = Integer.MAX_VALUE;
int populationSize =
10; // the number of samples for decision variables, one of them will be drawn randomly for
// the simulation.
boolean interpolate = true;
boolean includeCurrentBest = false;
// randomize the decision variables (for e.g.\ utility parameters for modes)
DecisionVariableRandomizer<ModeChoiceDecisionVariable> decisionVariableRandomizer =
new ModeChoiceRandomizer(scenario, RandomizedUtilityParametersChoser.ONLY_ASC, EQUIL);
// what would be the decision variables to optimize the objective function.
ModeChoiceDecisionVariable initialDecisionVariable =
new ModeChoiceDecisionVariable(scenario.getConfig().planCalcScore(), scenario, EQUIL);
// what would decide the convergence of the objective function
final int iterationsToConvergence = 200; //
final int averagingIterations = 10;
ConvergenceCriterion convergenceCriterion =
new FixedIterationNumberConvergenceCriterion(iterationsToConvergence, averagingIterations);
RandomSearch<ModeChoiceDecisionVariable> randomSearch =
new RandomSearch<>(
simulator,
decisionVariableRandomizer,
initialDecisionVariable,
convergenceCriterion,
maxIterations, // this many times simulator.run(...) and thus controler.run() will be
// called.
maxTransitions,
populationSize,
MatsimRandom.getRandom(),
interpolate,
objectiveFunction,
includeCurrentBest);
// probably, an object which decide about the inertia
SelfTuner selfTuner = new SelfTuner(0.95);
randomSearch.setLogPath(OUT_DIR);
// run it, this will eventually call simulator.run() and thus controler.run
randomSearch.run(selfTuner);
}
