/// <summary>
/// Initializes the agent. Can be safely called multiple times.
/// </summary>
public void LazyInitialize()
{
if (m_Initialized)
{
return;
}
m_Initialized = true;
// Grab the "static" properties for the Agent.
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_PolicyFactory = GetComponent <BehaviorParameters>();
m_Info = new AgentInfo();
m_Action = new AgentAction();
sensors = new List <ISensor>();
Academy.Instance.AgentIncrementStep += AgentIncrementStep;
Academy.Instance.AgentSendState += SendInfo;
Academy.Instance.DecideAction += DecideAction;
Academy.Instance.AgentAct += AgentStep;
Academy.Instance.AgentForceReset += _AgentReset;
m_Brain = m_PolicyFactory.GeneratePolicy(Heuristic);
ResetData();
Initialize();
InitializeSensors();
// The first time the Academy resets, all Agents in the scene will be
// forced to reset through the <see cref="AgentForceReset"/> event.
// To avoid the Agent resetting twice, the Agents will not begin their
// episode when initializing until after the Academy had its first reset.
if (Academy.Instance.TotalStepCount != 0)
{
OnEpisodeBegin();
}
}
void NotifyAgentDone(DoneReason doneReason)
{
m_Info.episodeId = m_EpisodeId;
m_Info.reward = m_Reward;
m_Info.done = true;
m_Info.maxStepReached = doneReason == DoneReason.MaxStepReached;
// Request the last decision with no callbacks
// We request a decision so Python knows the Agent is done immediately
m_Brain?.RequestDecision(m_Info, sensors);
// We also have to write any to any DemonstationStores so that they get the "done" flag.
foreach (var demoWriter in DemonstrationWriters)
{
demoWriter.Record(m_Info, sensors);
}
if (doneReason != DoneReason.Disabled)
{
// We don't want to udpate the reward stats when the Agent is disabled, because this will make
// the rewards look lower than they actually are during shutdown.
UpdateRewardStats();
}
// The Agent is done, so we give it a new episode Id
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_Reward = 0f;
m_CumulativeReward = 0f;
m_RequestAction = false;
m_RequestDecision = false;
}
/// <summary>
/// Initializes the agent. Can be safely called multiple times.
/// </summary>
public void LazyInitialize()
{
if (m_Initialized)
{
return;
}
m_Initialized = true;
// Grab the "static" properties for the Agent.
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_PolicyFactory = GetComponent <BehaviorParameters>();
m_Info = new AgentInfo();
m_Action = new AgentAction();
sensors = new List <ISensor>();
Academy.Instance.AgentIncrementStep += AgentIncrementStep;
Academy.Instance.AgentSendState += SendInfo;
Academy.Instance.DecideAction += DecideAction;
Academy.Instance.AgentAct += AgentStep;
Academy.Instance.AgentForceReset += _AgentReset;
m_Brain = m_PolicyFactory.GeneratePolicy(Heuristic);
ResetData();
InitializeAgent();
InitializeSensors();
}
/// MonoBehaviour function that is called when the attached GameObject
/// becomes enabled or active.
void OnEnable()
{
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
OnEnableHelper();
m_Recorder = GetComponent <DemonstrationRecorder>();
}
void NotifyAgentDone(bool maxStepReached = false)
{
m_Info.reward = m_Reward;
m_Info.done = true;
m_Info.maxStepReached = maxStepReached;
// Request the last decision with no callbacks
// We request a decision so Python knows the Agent is done immediately
m_Brain?.RequestDecision(m_Info, sensors, (a) => {});
// The Agent is done, so we give it a new episode Id
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_Reward = 0f;
m_CumulativeReward = 0f;
m_RequestAction = false;
m_RequestDecision = false;
}
void NotifyAgentDone(bool maxStepReached = false)
{
m_Info.reward = m_Reward;
m_Info.done = true;
m_Info.maxStepReached = maxStepReached;
// Request the last decision with no callbacks
// We request a decision so Python knows the Agent is done immediately
m_Brain?.RequestDecision(m_Info, sensors);
if (m_Recorder != null && m_Recorder.record && Application.isEditor)
{
m_Recorder.WriteExperience(m_Info, sensors);
}
UpdateRewardStats();
// The Agent is done, so we give it a new episode Id
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_Reward = 0f;
m_CumulativeReward = 0f;
m_RequestAction = false;
m_RequestDecision = false;
}
void NotifyAgentDone(bool maxStepReached = false)
{
m_Info.reward = m_Reward;
m_Info.done = true;
m_Info.maxStepReached = maxStepReached;
// Request the last decision with no callbacks
// We request a decision so Python knows the Agent is done immediately
m_Brain?.RequestDecision(m_Info, sensors);
// We also have to write any to any DemonstationStores so that they get the "done" flag.
foreach (var demoWriter in DemonstrationWriters)
{
demoWriter.Record(m_Info, sensors);
}
UpdateRewardStats();
// The Agent is done, so we give it a new episode Id
m_EpisodeId = EpisodeIdCounter.GetEpisodeId();
m_Reward = 0f;
m_CumulativeReward = 0f;
m_RequestAction = false;
m_RequestDecision = false;
}
