protected override void AddNewStateToVS(IState state)
{
if (!States.Contains(state))
{
States.Add(state);
}
V_s.Add(state, 0f);
}
public override int Think(IState state)
{
if (state.IsFinal)
{
return(0);
}
IState newState;
while (true)
{
float delta = 0;
foreach (var s in States.ToList())
{
if (s.IsFinal)
{
continue; // ne modifie pas V_s si état final
}
if (!V_s.ContainsKey(s))
{
AddNewStateToVS(s);
}
float tmp = V_s[s];
float max = -INFINITY;
foreach (var action in s.Actions)
{
float current = GetRewardForAction(s, action, out newState);
if (current > max)
{
max = current;
}
}
V_s[s] = max;
delta = Math.Max(delta, Math.Abs(tmp - V_s[s]));
}
if (delta < EPSILON)
{
break;
}
}
return(ArgMaxAction(state));
}
protected override void AddNewStateToVS(IState state)
{
if (!States.Contains(state))
{
States.Add(state);
}
// Intialise V_s avec un nombre aléatoire entre 0 et 1
float value = RandomFloat(0, 1);
V_s.Add(state, value);
if (state.IsFinal)
{
V_s[state] = 0; // Sauf si c'est un état final
}
}
// Met à jour V_s pour la stratégie courante.
private void PolicyEvaluation()
{
IState newState;
while (true)
{
float delta = 0;
foreach (var state in States.ToList())
{
if (state.IsFinal)
{
continue; // ne modifie pas V_s si état final
}
if (!V_s.ContainsKey(state))
{
AddNewStateToVS(state);
}
if (!policy.ContainsKey(state))
{
if (state.HasActions)
{
AddNewStateToPolicy(state);
}
else
{
continue; // skip if it's an no action state
}
}
float tmp = V_s[state];
V_s[state] = GetRewardForAction(state, (int)policy[state], out newState);
delta = Math.Max(delta, Math.Abs(tmp - V_s[state]));
}
if (delta < EPSILON)
{
break;
}
}
}