public void Evaluate(
IGridWorldPolicy policy,
IGridWorldRewarder gridWorldRewarder,
int sweepLimit = -1)
{
var sweep = 0;
var largestValueChange = 0.0;
do
{
largestValueChange = 0.0;
foreach (var state in _world.AllStates())
{
var originalValue = Value(state);
var newValue = CalculateValue(state, policy, gridWorldRewarder);
_values[state.Position1D] = newValue;
var valueChange = Math.Abs(originalValue - newValue);
if (valueChange > largestValueChange)
{
largestValueChange = valueChange;
}
}
if (sweepLimit > 0 && ++sweep == sweepLimit)
{
break;
}
} while (largestValueChange > 0.000001);
}
public static GreedyGridWorldPolicy Create(
GridWorld world,
GridWorldValueTable gridWorldValueTable,
IGridWorldRewarder gridWorldRewarder)
{
var greedyPolicy = new GreedyGridWorldPolicy();
foreach (var state in world.AllStates())
{
greedyPolicy._actions[state] = FindBestAction(world, state, gridWorldValueTable, gridWorldRewarder);
}
return(greedyPolicy);
}
private double CalculateValue(
GridWorldState state,
IGridWorldPolicy policy,
IGridWorldRewarder gridWorldRewarder)
{
var newValue = 0.0;
foreach (var action in _world.AvailableActions(state))
{
var nextState = _world.NextState(state, action);
var reward = gridWorldRewarder.Reward(state, action);
newValue += policy.PAction(state, action) * (reward + Value(nextState));
}
return(newValue);
}
private static GridWorldAction FindBestAction(
GridWorld world,
GridWorldState state,
GridWorldValueTable gridWorldValueTable,
IGridWorldRewarder gridWorldRewarder)
{
var max = double.MinValue;
var maxAction = GridWorldAction.Down;
foreach (var action in world.AvailableActions(state))
{
var nextState = world.NextState(state, action);
var nextStateValue = gridWorldValueTable.Value(nextState);
var reward = gridWorldRewarder.Reward(state, action);
if (reward + nextStateValue > max)
{
max = reward + nextStateValue;
maxAction = action;
}
}
return(maxAction);
}
