Merge pull request #73 from UoA-CARES/master

Update branch
UoA-CARES · Jul 16, 2023 · 223d19a · 223d19a
2 parents 9d0fd02 + bb333ae
commit 223d19a
Show file tree

Hide file tree

Showing 10 changed files with 222 additions and 195 deletions.
diff --git a/src/environments/environments/CarTrackEnvironment.py b/src/environments/environments/CarTrackEnvironment.py
@@ -160,7 +160,7 @@ def compute_reward(self, state, next_state):
             self.step_counter = 0
             self.update_goal_service(self.goal_number)
 
-        if has_collided(next_state[8:-2], self.COLLISION_RANGE) or has_flipped_over(next_state[2:6]):
+        if has_collided(next_state[8:], self.COLLISION_RANGE) or has_flipped_over(next_state[2:6]):
             reward -= 25  # TODO: find optimal value for this
 
         return reward
diff --git a/src/environments/environments/F1tenthEnvironment.py b/src/environments/environments/F1tenthEnvironment.py
@@ -30,7 +30,7 @@ def __init__(self, env_name, car_name, max_steps, step_length):
         self.STEP_LENGTH = step_length
 
         self.MAX_ACTIONS = np.asarray([3, 3.14])
-        self.MIN_ACTIONS = np.asarray([-0.5, -3.14])
+        self.MIN_ACTIONS = np.asarray([0, -3.14])
 
         self.ACTION_NUM = 2
 

diff --git a/src/environments/environments/util.py b/src/environments/environments/util.py
@@ -68,7 +68,7 @@ def avg_reduce_lidar(lidar: LaserScan):
 
 def reduce_lidar(lidar: LaserScan):
         ranges = lidar.ranges
-        ranges = np.nan_to_num(ranges, posinf=float(-1), neginf=float(-1))
+        ranges = np.nan_to_num(ranges, posinf=float(10), neginf=float(0))
         ranges = list(ranges)
 
         reduced_range = []

diff --git a/src/f1tenth b/src/f1tenth
diff --git a/src/reinforcement_learning/config/test.yaml b/src/reinforcement_learning/config/test.yaml
@@ -1,9 +1,10 @@
 test:
   ros__parameters:
-    environment: 'CarTrack1'
+    environment: 'CarTrack'
+    track: 'track_2'
     max_steps_evaluation: 1000000
-    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
+    actor_path: rl_logs/23_07_06_01:46:43/models/actor_checkpoint.pht
+    critic_path: rl_logs/23_07_06_01:46:43/models/critic_checkpoint.pht
     max_steps: 300
     step_length: 0.25
     reward_range: 2.0

diff --git a/src/reinforcement_learning/launch/sanity_check.launch.py b/src/reinforcement_learning/launch/sanity_check.launch.py
@@ -1,20 +1,37 @@
 import os
 from ament_index_python import get_package_share_directory
-from launch_ros.actions import Node 
+from launch_ros.actions import Node, SetParameter
 from launch import LaunchDescription
 from launch.launch_description_sources import PythonLaunchDescriptionSource
 from launch.actions import IncludeLaunchDescription, SetEnvironmentVariable
+from launch.substitutions import TextSubstitution
+import yaml
+
+env_launch = {
+    'CarGoal': 'cargoal',
+    'CarWall': 'carwall',
+    'CarBlock': 'carblock',
+    'CarTrack': 'cartrack'
+}
 
 def generate_launch_description():
     pkg_f1tenth_description = get_package_share_directory('f1tenth_description')
     pkg_f1tenth_bringup = get_package_share_directory('f1tenth_bringup')
     pkg_environments = get_package_share_directory('environments')
 
+    config_path = os.path.join(
+        get_package_share_directory('reinforcement_learning'),
+        'train.yaml'
+    )
+
+    config = yaml.load(open(config_path), Loader=yaml.Loader)
+    env = config['train']['ros__parameters']['environment']
+
     environment =  IncludeLaunchDescription(
         launch_description_source=PythonLaunchDescriptionSource(
-            os.path.join(pkg_environments, 'cargoal.launch.py')),
+            os.path.join(pkg_environments, f'{env_launch[env]}.launch.py')),
         launch_arguments={
-            'car_name': 'f1tenth',
+            'track': TextSubstitution(text=str(config['train']['ros__parameters']['track'])),
         }.items() #TODO: this doesn't do anything
     )
 
@@ -27,17 +44,12 @@ def generate_launch_description():
         }.items()
     )
 
-    config = os.path.join(
-        get_package_share_directory('reinforcement_learning'),
-        'train.yaml'
-    )
-
     # Launch the Environment
     main = Node(
             package='reinforcement_learning',
             executable='sanity_check',
             parameters=[
-                config
+                config_path
             ],
             name='sanity_check',
             output='screen',
@@ -47,6 +59,7 @@ def generate_launch_description():
     return LaunchDescription([
         #TODO: Find a way to remove this
         SetEnvironmentVariable(name='GZ_SIM_RESOURCE_PATH', value=pkg_f1tenth_description[:-19]),
+        SetParameter(name='use_sim_time', value=True),
         environment,
         f1tenth,
         main

diff --git a/src/reinforcement_learning/launch/test.launch.py b/src/reinforcement_learning/launch/test.launch.py
@@ -4,15 +4,14 @@
 from launch import LaunchDescription
 from launch.launch_description_sources import PythonLaunchDescriptionSource
 from launch.actions import IncludeLaunchDescription, SetEnvironmentVariable
+from launch.substitutions import TextSubstitution
 import yaml
 
 env_launch = {
     'CarGoal': 'cargoal',
     'CarWall': 'carwall',
     'CarBlock': 'carblock',
-    'CarTrack': 'cartrack',
-    'CarTrack1': 'cartrack1',
-    'CarTrack2': 'cartrack2'
+    'CarTrack': 'cartrack'
 }
 
 def generate_launch_description():
@@ -32,7 +31,7 @@ def generate_launch_description():
         launch_description_source=PythonLaunchDescriptionSource(
             os.path.join(pkg_environments, f'{env_launch[env]}.launch.py')),
         launch_arguments={
-            'car_name': 'f1tenth',
+            'track': TextSubstitution(text=str(config['test']['ros__parameters']['track'])),
         }.items() #TODO: this doesn't do anything
     )