private void PolicyIteration()
{
StateValueTable valueTable = new StateValueTable(m_initMode, m_initValue,
this.EnvRows, this.EnvCols, // my pos
this.EnvRows, this.EnvCols, // opp pos
2); // ball ownership status
m_valueTable = valueTable;
Policy policy = new Policy(m_initMode, m_initMode == ValueTableInitModes.Constant ? 0 : m_numActions,
this.EnvRows, this.EnvCols, // my pos
this.EnvRows, this.EnvCols, // opp pos
2); // ball ownership status
m_piPolicy = policy;
long stateCounts = m_valueTable.NumStates;
while (true)
{
// Policy Evaluation
double delta = 0.0;
do
{
delta = 0.0;
// foreach state
for (int s = 0; s < stateCounts; s++)
{
double v = valueTable.GetValueLinear(s);
double newV = EstimateNewValueUsingPolicy(s, policy);
valueTable.SetValueLinear(newV, s);
delta = Math.Max(delta, Math.Abs(v - newV));
}
} while (delta > Theta);
// Policy Improvement
bool policyStable = true;
for (int s = 0; s < stateCounts; s++)
{
int b = policy.GetValueLinear(s);
int bestAct;
EstimateNewValue((int)s, out bestAct);
policy.SetValueLinear(bestAct, s);
if (b != bestAct)
{
policyStable = false;
}
}
if (policyStable)
{
break;
}
}
}
private double EstimateNewValueUsingPolicy(int s, Policy policy)
{
StateValueTable valueTable = m_valueTable as StateValueTable;
if (valueTable == null)
{
throw new Exception("A ValueTable needed!");
}
List <int> nextStates;
List <double> nextStateProbs;
List <double> nextStateRew;
int a = policy.GetValueLinear(s);
GetPossibleNextStates(s, a, out nextStates, out nextStateProbs, out nextStateRew);
double nextV = 0.0;
for (int i = 0; i < nextStates.Count; i++)
{
nextV += nextStateProbs[i] * (nextStateRew[i] + (Gamma * valueTable.GetValueLinear(nextStates[i])));
}
return(nextV);
}
private void ValueIteration()
{
StateValueTable valueTable = new StateValueTable(m_initMode, m_initValue,
this.EnvRows, this.EnvCols, // my pos
this.EnvRows, this.EnvCols, // opp pos
2); // ball ownership status
m_valueTable = valueTable;
long stateCounts = m_valueTable.NumStates;
double delta = 0.0;
do
{
delta = 0.0;
// foreach state
for (int s = 0; s < stateCounts; s++)
{
double v = valueTable.GetValueLinear(s);
double newV = EstimateNewValue(s);
valueTable.SetValueLinear(newV, s);
delta = Math.Max(delta, Math.Abs(v - newV));
}
} while (delta > Theta);
}
private double EstimateNewValue(int s, out int maxA)
{
StateValueTable valueTable = m_valueTable as StateValueTable;
if (valueTable == null)
{
throw new Exception("A ValueTable needed!");
}
List <int> nextStates;
List <double> nextStateProbs;
List <double> nextStateRew;
double maxV = Double.MinValue;
maxA = -1;
for (int a = 0; a < m_numActions; a++)
{
GetPossibleNextStates(s, a, out nextStates, out nextStateProbs, out nextStateRew);
double nextV = 0.0;
for (int i = 0; i < nextStates.Count; i++)
{
nextV += nextStateProbs[i] * (nextStateRew[i] + (Gamma * valueTable.GetValueLinear(nextStates[i])));
}
if (nextV > maxV)
{
maxV = nextV;
maxA = a;
}
}
return(maxV);
}