rename rabbit_learn to Frozen_Lake

qlearn_0/src/Frozen_Lake.c

@@ -1,7 +1,6 @@
 
-#include "rabbit_learn.h"
+#include "Frozen_Lake.h"
 
-int ADD_MOVE[ACTION_COUNT]={ MOVE_DOWN, MOVE_LEFT, MOVE_RIGHT, MOVE_UP };
 char * action_name ="DLRU";
 char * content_name ="ESCBF";
 
@@ -35,6 +34,17 @@ struct game_params * create_game_params (
   gm_param->limit_EPISODES_number = limit_EPISODES_number;
   gm_param->limit_MOVE_number = limit_MOVE_number;
   
+  gm_param->add_move = malloc(ACTION_COUNT * sizeof(long int));
+/*
+  gm_param->add_move[DOWN] = signedLineFromCoord((long int []){0,-1}, dim);
+  gm_param->add_move[LEFT] = signedLineFromCoord((long int []){-1,0}, dim);
+  gm_param->add_move[RIGHT] = signedLineFromCoord((long int []){1,0}, dim);
+  gm_param->add_move[UP] = signedLineFromCoord((long int []){0,1}, dim);
+*/
+  gm_param->add_move[DOWN] = dim->perm[0];
+  gm_param->add_move[LEFT] = -1;
+  gm_param->add_move[RIGHT] = 1;
+  gm_param->add_move[UP] = -1*dim->perm[0];
   return gm_param;
 }
 
@@ -57,6 +67,7 @@ struct delay_params * create_delay_params (
 }
 
 void reset_game_status(struct game_status * status){
+  status->rabbitOldRankPosition = status->startRankPosition ;
   status->rabbitRankPosition = status->startRankPosition ;
   status->endGame = false;
   status->count_MOVES = 0;
@@ -82,11 +93,11 @@ struct game_status * create_game_status(){
 
 long int generate_game(struct game *gm){
   struct game_params *params = gm->params;
-  struct game_status *status;
+  struct game_status *status = gm->status;
   dimension *dim = params->dim;
   for(long int i=0; i<(params->dim)->rank; ++i){
     (gm->cells[i]).rankPosition = i;
-    (gm->cells[i]).content = 0;
+    (gm->cells[i]).content = EMPTY; //0
     for(long int j=0; j < ACTION_COUNT; ++j)
       (gm->cells[i]).Q[j] = 0;
   }
@@ -101,6 +112,7 @@ long int generate_game(struct game *gm){
   (gm->cells[random]).content = START;
   
   (status)->startRankPosition = random;
+  (status)->rabbitOldRankPosition = random;
   (status)->rabbitRankPosition = random;
 
   return (status->startRankPosition);
@@ -117,6 +129,8 @@ struct game * create_game(
   gm->cells = create_game_cells(params->dim);
   gm->delay = delay; // create_delay_params(10000000, 2000000);
   gm->status = create_game_status();
+
+
   generate_game(gm);
 
   return gm; 
@@ -133,10 +147,33 @@ void free_game(struct game *gm){
   }
   free(gm->cells);
   free_dimension((gm->params)->dim);
+  free((gm->params)->add_move);
   free(gm->params);
   free(gm->qlearnParams);
   free(gm->delay);
   free(gm->status);
+  free(gm);
+}
+
+long int function_add_move(struct game *gm, enum Action action){
+  struct game_status * status = gm->status;
+  dimension *dim = (gm->params)->dim;
+
+  if(action == UP && (status->rabbitRankPosition / dim->perm[0]) == 0){
+    return dim->rank * -1;
+  }
+  if(action == DOWN && (status->rabbitRankPosition / dim->perm[0]) == dim->perm[1]-1){
+    return dim->rank ;
+  }
+  if(action == LEFT && (status->rabbitRankPosition % dim->perm[0]) == 0){
+    return dim->rank * -1;
+  }
+  if(action == RIGHT && (status->rabbitRankPosition % dim->perm[0]) == dim->perm[1]-1){
+    return dim->rank;
+  }
+
+  return (gm->params)->add_move[action];
+  
 }
 
 void move_game(struct game *gm, enum Action action){
@@ -145,9 +182,11 @@ void move_game(struct game *gm, enum Action action){
   struct cell * cells = gm->cells;
   
   ++(status->count_MOVES);
-  long int newRankPosition = status->rabbitRankPosition + ADD_MOVE[action];
+  long int newRankPosition = status->rabbitRankPosition + function_add_move(gm, action);
 
-  if (newRankPosition < 0 || newRankPosition >= (params->dim)->rank) status->reward = REWARD_OUT;
+  if (newRankPosition < 0 || newRankPosition >= (params->dim)->rank) {
+    status->reward = REWARD_OUT;
+  }
   else if (status->count_MOVES > params->limit_MOVE_number) {
     status->rabbitRankPosition = newRankPosition;              
     status->endGame =  true;
@@ -252,10 +291,10 @@ void mainQlearning_game(struct game *gm){
   double proba_explor;
   srand(time(NULL));
 
-  for(size_t k=0 ; k < params->limit_game_number; ++k){
+  for(long int k=0 ; k < params->limit_game_number; ++k){
-    for(size_t episode = 0; episode < params->limit_EPISODES_number; ++episode){
-      reset_game_status(status);
       generate_game(gm);
+    for(long int episode = 0; episode < params->limit_EPISODES_number; ++episode){
+      reset_game_status(status);
 
       while(!(status->endGame)){
         random = rand() % NUMBER_EPISODE2;
@@ -269,18 +308,20 @@ void mainQlearning_game(struct game *gm){
           printf("greedy action ");
         }
         
+        status->rabbitOldRankPosition = status->rabbitRankPosition;
         move_game(gm, action);
 
-  printf("action = %d rbPos = %ld, rwds = %ld, final_rwrd=%ld\n",action, status->rabbitRankPosition,status->reward, status->final_reward);
+  printf("ik=%ld, episode = %ld , action = %d rbPos = %ld, rwds = %ld, final_rwrd=%ld\n",k,episode,action, status->rabbitRankPosition,status->reward, status->final_reward);
         // update Q array on the action of the state
-         cells[status->rabbitRankPosition].Q[action] =  cells[status->rabbitRankPosition].Q[action] + (qlearnParams->learning_rate * (status->reward + qlearnParams->discount_factor * MAX_ARRAY_TYPE_DOUBLE(cells[status->rabbitRankPosition].Q, ACTION_COUNT) -         cells[status->rabbitRankPosition].Q[action]));
+         cells[status->rabbitOldRankPosition].Q[action] =  cells[status->rabbitOldRankPosition].Q[action] + (qlearnParams->learning_rate * (status->reward + qlearnParams->discount_factor * MAX_ARRAY_TYPE_DOUBLE(cells[status->rabbitRankPosition].Q, ACTION_COUNT) -         cells[status->rabbitOldRankPosition].Q[action]));
+         //cells[status->rabbitRankPosition].Q[action] =  cells[status->rabbitRankPosition].Q[action] + (qlearnParams->learning_rate * (status->reward + qlearnParams->discount_factor * MAX_ARRAY_TYPE_DOUBLE(cells[status->rabbitRankPosition].Q, ACTION_COUNT) -         cells[status->rabbitRankPosition].Q[action]));
         //cells[status->rabbitRankPosition].Q[action] =  (1 - qlearnParams->learning_rate) * cells[status->rabbitRankPosition].Q[action] + (qlearnParams->learning_rate * (status->reward + qlearnParams->discount_factor * MAX_ARRAY_TYPE_DOUBLE(cells[status->rabbitRankPosition].Q, ACTION_COUNT )));
          
         status->final_reward += status->reward;
 
         print_game_dim2(gm);
       
-        usleep((gm->delay)->delay_between_episodes);
+  //      usleep((gm->delay)->delay_between_episodes);
         
       }
       push_back_list_TYPE_L_INT(list_final_rewards, status->final_reward);
@@ -290,7 +331,7 @@ void mainQlearning_game(struct game *gm){
       printf(" %ld ",(list_final_rewards->current_list)->value);
     }
     remove_all_list_in_TYPE_L_INT(list_final_rewards);
-    usleep((gm->delay)->delay_between_games);
+ //   usleep((gm->delay)->delay_between_games);
   }
   free(list_final_rewards);
 }

qlearn_0/src/Frozen_Lake.h

@@ -11,7 +11,7 @@
 #define ACTION_COUNT 4
 /* Down(0), Left(1), Right(2), Up(3) */
 enum Action { DOWN, LEFT, RIGHT, UP };
-enum MoveAction { MOVE_DOWN = -2, MOVE_LEFT = -1, MOVE_RIGHT = 1, MOVE_UP = 2 };
+//enum MoveAction { MOVE_DOWN = -2, MOVE_LEFT = -1, MOVE_RIGHT = 1, MOVE_UP = 2 };
 
 #define CONTENT_COUNT 5 
 /* Empty(0), Start(1), Carrot(2), Block(3), Fox(4) */
@@ -22,6 +22,7 @@ enum Reward { REWARD_MOVES_OUT=-10, REWARD_OUT=-1, REWARD_EMPTY=-1, REWARD_CARRO
 struct game_params {
   size_t limit_game_number;
   dimension *dim;
+  long int *add_move;
   size_t limit_FOX_number;
   size_t limit_BLOCK_number;
   size_t limit_CARROT_number;
@@ -37,6 +38,7 @@ struct delay_params {
 struct game_status {
   long int startRankPosition;
   long int rabbitRankPosition;
+  long int rabbitOldRankPosition;
   bool endGame;
   long int reward;
   long int final_reward;

qlearn_0/test/Makefile

@@ -0,0 +1,75 @@
+
+
+
+
+NAME_TEST=is_good
+CC=gcc
+ROOT_DIR=$(PWD)
+FROZENLAKEDIR=$(PWD)/..
+YTESTDIR=$(PWD)/../../ytest_t
+YPERMDIR=$(PWD)/../../ypermutation_t
+DIMDIR=$(PWD)/../../dimension_t
+LISTDIR=$(PWD)/../../list_t
+
+INCLUDE_DIR=$(PWD)/../src
+CFLAGS=-I$(INCLUDE_DIR) -I$(YTESTDIR)/include_ytest/include  -I$(YPERMDIR)/src   -I$(DIMDIR)/src -I$(LISTDIR)/src
+LDFLAGS=-L$(YTESTDIR) -lytest #"-D DEBUG=1"
+
+#SRC_DIR=$(ROOT_DIR)/src
+#SRC=$(wildcard */*/*.c) 
+SRC=$(wildcard **.c) 
+#HEADS=$(OBJS:.o=.h)
+TEST_DIR=$(PWD)
+EXECSRC=$(NAME_TEST).c
+EXEC=launch_$(NAME_TEST)_m
+LISTSRC_O=$(LISTDIR)/src/list_t/list_t.o
+
+FROZENLAKESRC=$(FROZENLAKEDIR)/src/Frozen_Lake.c
+FROZENLAKESRC_O=$(FROZENLAKESRC:.c=.o)
+
+
+PERMSRC_O=$(YPERMDIR)/src/permutation_t/permutation_t.o
+DIMSRC_O=$(DIMDIR)/src/dimension_t/dimension_t.o
+
+
+TOPTARGETS := all clean 
+
+DEPS=$(PERMSRC) $(DIMDIR) $(LISTDIR) $(YTESTDIR)
+
+$(TOPTARGETS): $(DEPS) 
+
+$(DEPS):
+	$(MAKE) -C $@ $(MAKECMDGOALS)
+
+
+#LISTSRC_O=$(LISTSRC:.c=.o)
+#SETTSRC_O=$(PWD)/../src/set_theoric_t/set_theoric_t.o
+#SETTSRC_O=$(SETTSRC:.c=.o)
+#TOOLSRC=$(TOOLDIR)/src/tools_t/tools_t.c
+#TOOLSRC_O=$(TOOLSRC:.c=.o)
+
+#OBJ=$(SRC:.c=.o)  $(FROZENLAKESRC_O)
+OBJ=$(FROZENLAKESRC_O) $(LISTSRC_O) $(PERMSRC_O)  $(DIMSRC_O)
+
+LIB_YTEST=$(YTESTDIR)/libytest.so
+
+all: $(EXEC) $(LIB_YTEST)
+
+$(EXEC):	$(EXECSRC) $(OBJ)
+	$(CC) -o $@ $^ $(CFLAGS) $(LDFLAGS)
+
+$(FROZENLAKESRC_O): $(FROZENLAKESRC) $(LISTSRC_O) $(PERMSRC_O)  $(DIMSRC_O)
+	$(CC) -o $@ -c $^ $(CFLAGS) 
+  
+#$(LDFLAGS)
+
+.PHONY: clean mrproper
+
+clean:
+	rm -f $(OBJ)
+
+mrproper: clean
+	rm -f $(EXEC)
+
+run: $(EXEC)
+	$(EXEC) -h

qlearn_0/test/compile.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+
+if [ "$#" -le 0 ] ; then
+  echo "Usage: $0 is_good.c" >&2
+  echo "for example to compile: is_good.c" >&2
+  exit 1
+fi
+if [ "$#" -le 1 ] ; then
+  echo "Usage: $0 $1" >&2
+  echo " we can add more option for example '-D DEBUG=1' to have debug print, '-D HK' to have gtest like prompt, od '-g' to gbd" >&2
+  echo "for example: $0 $1 \"-D DEBUG=1 -D HK -g\""
+fi
+
+DIR_YTEST=$PWD/../../ytest_t
+SRC=../src
+
+gcc -o launch_is_good_c $1 -L$DIR_YTEST $2 -lytest -I$DIR_YTEST/include_ytest/include $SRC/list_t/list_t.c -I$SRC
+#gcc -o launch_is_good_c $1 $2 -lytest -I../include_ytest src/list_t/list_t.o src/set_theoric_t/set_theoric_t.o -I./src 
+
+export LD_LIBRARY_PATH=$DIR_YTEST/:LD_LIBRARY_PATH
+
+
+#gcc $1 src/ftest/ftest.c src/fmock/fmock.c src/tools_t/tools_t.c src/bar_progress/bar_progress.c src/list_t/list_t.c src/set_theoric_t/set_theoric_t.c -I./include $2 -o launch_is_good_c -lpthread 

qlearn_0/test/is_good.c

@@ -0,0 +1,48 @@
+#include "ftest/ftest.h"
+#include "fmock/fmock.h"
+
+
+#include "Frozen_Lake.h"
+
+
+TEST(igameRabbit ){
+  size_t array[] = {4,4} ;
+  dimension *dim = init_copy_dim(array,2);
+  struct game_params * params = create_game_params(1,dim,3,3,1,200,200);
+  struct qlearning_params * qlearnParams = create_qlearning_params(0.85,0.99,1);
+  struct delay_params * delay_game = create_delay_params(1000000, 200000);
+  struct game * gm = create_game(params, qlearnParams, delay_game);
+
+  printDebug_dimension(gm->params->dim,"dimension game");
+  mainQlearning_game(gm);
+
+  free_game(gm);
+}
+
+
+
+/*
+TEST(ARG_MAX_ARRAY){
+
+  double Q[6] = {0.5,-2.3,-9,3,21,-65};
+  size_t mx = ARG_MAX_ARRAY_TYPE_DOUBLE(Q, 6) ;
+  size_t mn = ARG_MIN_ARRAY_TYPE_DOUBLE(Q, 6) ;
+  printf("mx = %ld, Q[%ld ] = %lf \n", mx,mx,Q[mx]);
+  printf("mn = %ld, Q[%ld ] = %lf \n", mn,mn,Q[mn]);
+
+  double a =-5;
+  double b=4; 
+
+  printf(" %lf vs %lf : %d \n ",a,b,COMPARE_N_TYPE_DOUBLE(&a,&b));
+  printf(" %lf vs %lf : %d \n ",b,a,COMPARE_N_TYPE_DOUBLE(&b,&a));
+
+}
+*/
+
+int main(int argc, char **argv){
+  
+
+  run_all_tests_args(argc, argv);
+
+  return 0;
+}