wip deepQlearning

deepQlearn_0/src/deepQlearning/learn_to_drive.c

@@ -71,6 +71,7 @@ struct delay_params * create_delay_params (
 }
 
 struct qlearning_params * create_qlearning_params  (
+  double gamma,
   double learning_rate,
   double discount_factor,
   double exploration_factor,
@@ -78,6 +79,7 @@ struct qlearning_params * create_qlearning_params  (
 ){
   struct qlearning_params * qparams = malloc(sizeof(struct qlearning_params));
 
+  qparams->gamma = gamma;
   qparams->learning_rate = learning_rate ;
   qparams->discount_factor = discount_factor ;
   qparams->exploration_factor = exploration_factor ;
@@ -148,10 +150,12 @@ void train_qlearning(struct RL_agent * rlAgent,
   tensor_TYPE_FLOAT * experimental_values = CREATE_TENSOR_FROM_CPY_DIM_TYPE_FLOAT(action_value->dim);
   
   struct game_status * car_status = rlAgent->car->status;
-  if( copy_tensor_TYPE_FLOAT(experimental_values, action_value) == 0 /* done */){
-    if(car_status->done){
-
-    }
+  copy_tensor_TYPE_FLOAT(experimental_values, action_value) ;
+  // experimental_values === Q-tab learning
+  if(car_status->done){
+    experimental_values->x[action] = -100;    
+  }else {
+    experimental_values->x[action] = reward + rlAgent->qlearnParams->gamma * MAX_ARRAY_TYPE_FLOAT(next_action_value->x, next_action_value->dim->rank) ;    
   }
   
 }

deepQlearn_0/src/deepQlearning/learn_to_drive.h

@@ -16,6 +16,7 @@
 
 
 struct qlearning_params {
+  double gamma;
   double learning_rate;
   double factor_update_learning_rate;
   double minimum_threshold_learning_rate;
@@ -66,6 +67,7 @@ struct delay_params * create_delay_params (
 );
 
 struct qlearning_params * create_qlearning_params (
+  double gamma,
   double learning_rate,
   double discount_factor,
   double exploration_factor,

deepQlearn_0/src/deepQlearning/vehicle.c

@@ -16,7 +16,7 @@
 
 struct game_status * create_game_status(){
   struct game_status * status = malloc(sizeof(struct game_status));
-  status->state = 0;
+  status->q_state = 0;
   status->reward = 0;
   status->cumulative_reward = 0;
   status->done = false;
@@ -482,7 +482,7 @@ void step(struct vehicle *v, int action){
   move_vehicle(v);
   read_sensor(v);
   struct game_status *status = v->status;
-  status->state =  v->sensor->x[LEFT]* 2500 + 
+  status->q_state =  v->sensor->x[LEFT]* 2500 + 
               v->sensor->x[CENTER]* 50 + 
               v->sensor->x[RIGHT] ;
   status->reward = 0;

deepQlearn_0/src/deepQlearning/vehicle.h

@@ -28,7 +28,7 @@
 
 
 struct game_status {
-  long state;
+  long q_state;
   long reward;
   long cumulative_reward;
   bool done;