//this is a temp depedency, until we get the feature extraction module
public Orchestrator(ITemperatureSensor temperatureProxy, IModelTraining trainerProxy)
{
temperatureProxy.Temperature.Subscribe(this, async signal =>
{
OrchestratorTwin twin;
lock (_twinLock)
{
twin = _lastTwin;
}
if (twin == null)
{
return(MessageResult.Ok);
}
if (twin.Scale == TemperatureScale.Celsius)
{
signal.Value = signal.Value * 9 / 5 + 32;
}
await BroadcastMessage(signal, twin);
return(MessageResult.Ok);
});
trainerProxy.Model.Subscribe(this, async model =>
{
Logger.LogInformation(
$"New Trained Model for {model.Algorithm}. Updating the model in {model.Algorithm}");
await Model.PublishAsync(model);
return(MessageResult.Ok);
});
Twin.Subscribe(async twin =>
{
lock (_twinLock)
{
_lastTwin = twin;
}
Logger.LogInformation($"Twin update. Routing = {twin.RoutingMode.ToString()}");
await Twin.ReportAsync(twin);
return(TwinResult.Ok);
});
}
public AnomalyDetection(IOrchestrator orcherstratorProxy, IModelTraining modelTrainingProxy)
{
orcherstratorProxy.Detection.Subscribe(this, async signal =>
{
try
{
int cluster = 0;
if (_kMeansScoring != null)
{
lock (_syncClustering)
if (_kMeansScoring != null)
{
cluster = _kMeansScoring.Score(new double[] { signal.Value });
}
}
if (cluster < 0)
{
await Anomaly.PublishAsync(new Anomaly()
{
Temperature = signal
});
}
}
catch (Exception ex)
{
Console.WriteLine(ex);
}
return(MessageResult.Ok);
});
modelTrainingProxy.Model.Subscribe(this, async(model) =>
{
//if the messages has been stored and forwarded, but the file has been deleted (e.g. a restart)
//then the message can be empty (null)
if (model == null)
{
return(MessageResult.Ok);
}
try
{
lock (_syncClustering)
switch (model.Algorithm)
{
case ThermostatApplication.Algorithm.kMeans:
_kMeansScoring = new KMeansScoring(_numClusters);
_kMeansScoring.DeserializeModel(model.DataJson);
break;
case ThermostatApplication.Algorithm.LSTM:
break;
default:
break;
}
}
catch (Exception ex)
{
Console.WriteLine(ex);
}
return(MessageResult.Ok);
});
}
