package rddl.solver.mdp.mcts;
/**
* This class will be used for the competitive portion of the project.
* Implement the best working selection and backpropate methods and explain in my_mcts.txt.
*
* Note: we’ll use the Elevators domain with 2 elevators and 5 floors for the evaluation.
*/
public class CompetitiveMCTS extends MCTS {
public CompetitiveMCTS(String instance_name) {
super(instance_name);
}
/**
* Returns the utility of the given StateActionNode for action selection.
*/
@Override
public double evaluateStateActionNode(StateActionNode node, boolean greedy) {
/**
* TODO: implement your chosen tree policy
*/
return 0;
}
/**
* Backpropagates the cumulative reward from a rollout trajectory to ancestral nodes.
*
* @param node the currently updated node (initially, a leaf node)
* @param cumRewardFromLeaf the cumulative reward to backpropagate (initially, the cumReward from simulation)
*/
@Override
public void backPropagate(TreeNode node, double cumRewardFromLeaf) {
/**
* TODO: implement your chosen back-up strategy
*/
}
}