/// <summary>
/// called after step and when the enviorment is resolved. return whether the enviourment should reset
/// </summary>
/// <param name="environment"></param>
public bool Record(IRLEnvironment environment)
{
Debug.Assert(environment.IsResolved());
bool isEnd = environment.IsEnd();
for (int i = 0; i < NumberOfActor; ++i)
{
float reward = environment.LastReward();
AddHistory(LastState[i], reward, LastAction[i], LastActionProbs[i], LastValue[i], i);
}
if (isEnd || environment.CurrentStep() >= MaxStepHorizon)
{
float[] nextValues = new float[NumberOfActor];
if (!isEnd)
{
nextValues = Model.EvaluateValue(environment.CurrentState());
}
else
{
for (int i = 0; i < NumberOfActor; ++i)
{
nextValues[i] = 0;
}
}
for (int i = 0; i < NumberOfActor; ++i)
{
ProcessEpisodeHistory(nextValues[i], i);
}
return(true);
}
return(false);
}
/// <summary>
/// called after step and when the enviorment is resolved. return whether the enviourment should reset
/// </summary>
/// <param name="environment"></param>
public virtual bool Record(IRLEnvironment environment)
{
Debug.Assert(environment.IsResolved());
bool isEnd = environment.IsEnd();
for (int i = 0; i < NumberOfActor; ++i)
{
float reward = environment.LastReward(i);
AddHistory(i, LastState[i], reward, LastAction[i], isEnd);
}
if (isEnd || environment.CurrentStep() >= MaxStepHorizon)
{
for (int i = 0; i < NumberOfActor; ++i)
{
UpdateReplayBuffer(i);
}
return(true);
}
return(false);
}
