public void SaveStateReward(
string previousAgentState,
string currentAgentState,
double reward,
TransportType previousAction)
{
var previousQFunc = db.QFuncInfos.FirstOrDefault(x => x.State == previousAgentState);
if (previousQFunc == null)
{
db.Add(StorageHelpers.CreateRandomQFuncInfo(previousAgentState));
}
else
{
var currenQFunc = db.QFuncInfos.FirstOrDefault(x => x.State == currentAgentState);
if (currenQFunc == null)
{
currenQFunc = StorageHelpers.CreateRandomQFuncInfo(currentAgentState);
db.Add(currenQFunc);
}
var maxNextReward = currenQFunc.GetBestReward();
if (previousAction == TransportType.Bus)
{
previousQFunc.BusReward = QLearningAlgoritm.GetUpdateReward(previousQFunc.BusReward, maxNextReward, reward);
}
else
{
previousQFunc.CarReward = QLearningAlgoritm.GetUpdateReward(previousQFunc.CarReward, maxNextReward, reward);
}
}
db.SaveChanges();
}
public TransportType GetBestNextTransport(string currentAgentState)
{
var qFuncInfo = qFuncInfos.FirstOrDefault(x => x.State == currentAgentState);
if (qFuncInfo == null)
{
qFuncInfo = StorageHelpers.CreateRandomQFuncInfo(currentAgentState);
qFuncInfos.Add(qFuncInfo);
}
return(QLearningAlgoritm.GetBestNextTransportWithEpsilonMush(qFuncInfo));
}
public void Should_rigth_update_reward(double previousReward, double maxNextReward, double currentReward, double expectedValue)
{
var result = QLearningAlgoritm.GetUpdateReward(previousReward, maxNextReward, currentReward);
result.Should().Be(expectedValue);
}
