PO_ETUCT.hh

Go to the documentation of this file.

#ifndef _PO_ETUCT_HH_
#define _PO_ETUCT_HH_

#include <rl_common/Random.h>
#include <rl_common/core.hh>

#include "../Models/FactoredModel.hh"

#include <set>
#include <vector>
#include <map>
#include <deque>

class PO_ETUCT: public Planner {
public:

  PO_ETUCT(int numactions, float gamma, float rrange, float lambda,
           int MAX_ITER, float MAX_TIME, int MAX_DEPTH,  int modelType,
           const std::vector<float> &featmax, const std::vector<float> &featmin,
           const std::vector<int> &statesPerDim, bool trackActual, 
           int history, Random rng = Random());
  
  PO_ETUCT(const PO_ETUCT &);

  virtual ~PO_ETUCT();

  virtual void setModel(MDPModel* model);
  virtual bool updateModelWithExperience(const std::vector<float> &last, 
                                         int act, 
                                         const std::vector<float> &curr, 
                                         float reward, bool term);
  virtual void planOnNewModel();
  virtual int getBestAction(const std::vector<float> &s);

  virtual void setSeeding(bool seed);
  virtual void setFirst();

  void logValues(ofstream *of, int xmin, int xmax, int ymin, int ymax);
  
  std::vector<float> discretizeState(const std::vector<float> &s);

  bool PLANNERDEBUG;
  bool MODELDEBUG;
  bool ACTDEBUG;
  bool UCTDEBUG;
  bool REALSTATEDEBUG;
  bool HISTORYDEBUG;

  MDPModel* model;

  typedef const std::vector<float> *state_t;


protected:


  struct state_info;
  struct model_info;

  struct state_samples {
    std::vector<state_t> samples;
  };

  struct state_info {

    // data filled in from models
    StateActionInfo* model;

    // q values from policy creation
    std::vector<float> Q;

    // uct experience data
    int uctVisits;
    std::vector<int> uctActions;
    short unsigned int visited;
    short unsigned int id;

    // needs update
    bool needsUpdate;

  };


  void initStateInfo(state_t s, state_info* info);
  
  state_t canonicalize(const std::vector<float> &s);

  void deleteInfo(state_info* info);
  
  void initNewState(state_t s);
  
  void createPolicy();
  
  void printStates();
  
  void calculateReachableStates();
  
  void removeUnreachableStates();

  void updateStateActionFromModel(state_t s, int a, state_info* info);

  void updateStateActionHistoryFromModel(const std::vector<float> modState, int a, StateActionInfo *newModel);

  double getSeconds();

  void resetUCTCounts();
  
  float uctSearch(const std::vector<float> &actualS, state_t state, int depth);
  
  std::vector<float> simulateNextState(const std::vector<float> &actualState, state_t discState, state_info* info, int action, float* reward, bool* term);
  
  int selectUCTAction(state_info* info);
  
  void canonNextStates(StateActionInfo* modelInfo);
  
  virtual void savePolicy(const char* filename);
  
  std::vector<float> addVec(const std::vector<float> &a, const std::vector<float> &b);
  
  std::vector<float> subVec(const std::vector<float> &a, const std::vector<float> &b);

private:

  std::set<std::vector<float> > statespace;

  std::map<state_t, state_info> statedata;

  std::deque<float> saHistory;

  std::vector<float> featmax;
  std::vector<float> featmin;
  
  state_t prevstate;
  int prevact;
  state_info* previnfo;

  double planTime;

  bool seedMode;

  int nstates;
  int nactions; 
  int lastUpdate;
  bool timingType;

  const int numactions;
  const float gamma;
  const float rrange;
  const float lambda;

  const int MAX_ITER;
  const float MAX_TIME;
  const int MAX_DEPTH;
  const int modelType;
  const std::vector<int> &statesPerDim;
  const bool trackActual;
  const int HISTORY_SIZE;
  const int HISTORY_FL_SIZE;

};

#endif