CLAW Library (a C++ Library Absolutely Wonderful) 1.5.5
Public Types | Public Member Functions

claw::ai::game::select_action< Method > Class Template Reference

Select an action using a given method (min_max, alpha_beta). More...

#include <game_ai.hpp>

List of all members.

Public Types

typedef Method::state state
typedef Method::action action
typedef Method::score score

Public Member Functions

void operator() (int depth, const state &current_state, action &new_action, bool computer_turn) const
 Select an action using the given method.

Detailed Description

template<typename Method>
class claw::ai::game::select_action< Method >

Select an action using a given method (min_max, alpha_beta).

Template parameters:

Author:
Julien Jorge, Sébastien Angibaud

Definition at line 188 of file game_ai.hpp.


Member Typedef Documentation

template<typename Method >
typedef Method::action claw::ai::game::select_action< Method >::action

Definition at line 192 of file game_ai.hpp.

template<typename Method >
typedef Method::score claw::ai::game::select_action< Method >::score

Definition at line 193 of file game_ai.hpp.

template<typename Method >
typedef Method::state claw::ai::game::select_action< Method >::state

Definition at line 191 of file game_ai.hpp.


Member Function Documentation

template<typename Method >
void claw::ai::game::select_action< Method >::operator() ( int  depth,
const state current_state,
action new_action,
bool  computer_turn 
) const

Select an action using the given method.

Parameters:
depthMaximum depth of the search tree.
current_stateThe state of the game.
new_action(in/out) Best known action.
computer_turnTell if the action is done by the computer.

Definition at line 328 of file game_ai.tpp.

{
  std::list<action> l;
  typename std::list<action>::iterator it;
  score best_eval;              
  Method method;

  // get all reachable states
  current_state.next_actions( l );
  best_eval = current_state.min_score();

  for (it=l.begin(); it!=l.end(); ++it)
    {
      state* new_state;
      score eval;
                        
      // try and evaluate each action
      new_state = static_cast<state*>(current_state.do_action(*it));
      eval = method(depth-1, *new_state, !computer_turn);

      delete new_state;

      // we keep one of the best actions
      if (eval > best_eval)
        {
          best_eval = eval;
          new_action = *it;
        }
    }
} // select_action::operator()

The documentation for this class was generated from the following files: