Merge pull request #64 from UoA-CARES/Experiment3-merged-12

Update f1tenth and lidar
UoA-CARES · Jul 10, 2023 · bdba878 · bdba878
2 parents f132707 + 47472b4
commit bdba878
Show file tree

Hide file tree

Showing 5 changed files with 11 additions and 9 deletions.
diff --git a/src/environments/environments/CarTrackParentEnvironment.py b/src/environments/environments/CarTrackParentEnvironment.py
@@ -40,7 +40,6 @@ def __init__(self, car_name, reward_range=1, max_steps=50, collision_range=0.2,
 
         # Environment Details ----------------------------------------
         self.MAX_STEPS_PER_GOAL = max_steps
-        self.MIN_ACTIONS = np.asarray([0, -3.14])
         self.OBSERVATION_SIZE = 8 + 10  # Car position + Lidar rays
 
         self.check_goal = False
@@ -151,8 +150,11 @@ def compute_reward(self, state, next_state):
 
         goal_position = self.goal_position
 
+        prev_distance = math.dist(goal_position, state[:2])
         current_distance = math.dist(goal_position, next_state[:2])
-
+
+        reward += prev_distance - current_distance
+
         if current_distance < self.REWARD_RANGE:
             reward += 50
             self.goal_number += 1

diff --git a/src/environments/environments/ParentCarEnvironment.py b/src/environments/environments/ParentCarEnvironment.py
@@ -49,7 +49,7 @@ def __init__(self, env_name, car_name, reward_range, max_steps, collision_range,
         self.lidar_sub = Subscriber(
             self,
             LaserScan,
-            f'/lidar',
+            f'/{self.NAME}/scan',
         )
 
         self.message_filter = ApproximateTimeSynchronizer(

diff --git a/src/f1tenth b/src/f1tenth
diff --git a/src/reinforcement_learning/config/test.yaml b/src/reinforcement_learning/config/test.yaml
@@ -2,9 +2,9 @@ test:
   ros__parameters:
     environment: 'CarTrack1'
     max_steps_evaluation: 1000000
-    actor_path: models/cartrack_1_training-26-06-2023-06:50:49_150000_actor.pht
-    critic_path: models/cartrack_1_training-26-06-2023-06:50:49_150000_critic.pht
-    max_steps: 100
+    actor_path: models/23_07_05_03:52:58/models/actor_checkpoint.pht
+    critic_path: models/23_07_05_03:52:58/models/critic_checkpoint.pht
+    max_steps: 300
     step_length: 0.25
-    reward_range: 0.2
+    reward_range: 2.0
     collision_range: 0.2
diff --git a/src/reinforcement_learning/reinforcement_learning/train.py b/src/reinforcement_learning/reinforcement_learning/train.py
@@ -110,7 +110,7 @@ def main():
         'reward_range': REWARD_RANGE,
         'collision_range': COLLISION_RANGE
     }
-    record = Record(networks=networks, checkpoint_freq=MAX_STEPS_TRAINING / 10, config=config)
+    record = Record(networks=networks, checkpoint_freq=100, config=config)
 
     train(env=env, agent=agent, record=record)
+23 −3		f1tenth_bringup/launch/simulation_bringup.launch.py
+2 −0		f1tenth_bringup/package.xml
+38 −0		f1tenth_description/config/ekf.yaml
+73 −0		f1tenth_description/config/slam_toolbox.yaml
+3 −2		f1tenth_description/gazebo/lidar.gazebo.xacro
+4 −0		f1tenth_description/gazebo/odometry.gazebo.xacro