Change folder name from result files to logs, add speed comparison fo…

…r intersection calculation

.gitignore

@@ -157,3 +157,4 @@ fabric.properties
 gym_dockauv/notes.py
 /gym_dockauv/tests/test_notes.py
 result_files/
+logs/

gym_dockauv/config/env_default_config.py

@@ -18,7 +18,7 @@
  # ---------- SIMULATION --------------
  "t_step_size": 0.10, # Length of each simulation timestep [s]
  "interval_datastorage": 20, # Interval of episodes on which extended data is saved through data class
- "save_path_folder": os.path.join(os.getcwd(), "result_files"), # Folder name where all result files will be stored
+ "save_path_folder": os.path.join(os.getcwd(), "logs"), # Folder name where all result files will be stored
 
  # ---------- GOAL ----------
  "goal_location": np.array([0, 0, 0]),

gym_dockauv/train.py

@@ -19,7 +19,7 @@ def train(total_timesteps: int) -> None:
  model = PPO('MlpPolicy', env, verbose=0)
  model.learn(total_timesteps=total_timesteps) # Train the agent
  # Save the agent
- model.save("PPO_docking")
+ model.save("logs/PPO_docking")
  env.save()
 
 
@@ -29,7 +29,7 @@ def predict():
  # NOTE: if you have loading issue, you can pass `print_system_info=True`
  # to compare the system on which the model was trained vs the current one
  # model = DQN.load("dqn_lunar", env=env, print_system_info=True)
- model = PPO.load("PPO_docking", env=env)
+ model = PPO.load("logs/PPO_docking", env=env)
 
  # Evaluate the agent
  # NOTE: If you use wrappers with your environment that modify rewards,
@@ -46,21 +46,21 @@ def predict():
  env.render()
 
 
-def post_analysis_directory():
- directory = "/home/erikx3/PycharmProjects/gym_dockauv/result_files"
+def post_analysis_directory(directory: str = "/home/erikx3/PycharmProjects/gym_dockauv/logs"):
 
  for file in os.listdir(directory):
  filename = os.fsdecode(file)
  # Capture full data pkl file
  full_path = os.path.join(directory, filename)
- if filename.endswith("FULL.pkl"):
+ if filename.endswith("FULL__DATA__STORAGE.pkl"):
  full_stor = FullDataStorage()
  full_stor.load(full_path)
  full_stor.plot_rewards()
+ # Episode Data Storage:
  elif filename.endswith(".pkl"):
  epi_stor = EpisodeDataStorage()
  epi_stor.load(full_path)
  epi_stor.plot_epsiode_states_and_u()
  epi_stor.plot_rewards()
  plt.show()
- #epi_stor.plot_episode_animation(t_per_step=None, title="Test Post Flight Visualization")
+ #epi_stor.plot_episode_animation(t_per_step=None, title="Test Post Flight Visualization")

gym_dockauv/utils/datastorage.py

@@ -35,7 +35,7 @@ def set_up_episode_storage(self, env, path_folder: str, title: str = "") -> None
  r"""
  Set up the storage to save and update incoming data with passing a reference of the env
 
- file_save_name: will be formatted path_folder\YYYY-MM-DDTHH-MM-SS__{title}__FULL.pkl
+ file_save_name: will be formatted path_folder\YYYY-MM-DDTHH-MM-SS__{title}__FULL_DATA_STORAGE.pkl
 
  :param env: The gym environment
  :param path_folder: Path to folder
@@ -49,7 +49,7 @@ def set_up_episode_storage(self, env, path_folder: str, title: str = "") -> None
  utc_str = datetime.datetime.utcnow().strftime('%Y_%m_%dT%H_%M_%S')
  if len(path_folder) > 0:
  os.makedirs(path_folder, exist_ok=True) # Create folder if not exists yet
- self.file_save_name = os.path.join(path_folder, f"{utc_str}__{title}__FULL.pkl")
+ self.file_save_name = os.path.join(path_folder, f"{utc_str}__{title}__FULL_DATA_STORAGE.pkl")
 
  cum_rewards_arr = ArrayList(env.cum_reward_arr)
  rewards_arr = ArrayList(env.last_reward_arr)
@@ -213,7 +213,7 @@ def set_up_episode_storage(self, path_folder: str, vehicle: AUVSim, step_size: f
  Set up the storage to save and update incoming data, including passing a reference to the vehicle and
  environment
 
- file_save_name: will be formatted path_folder\YYYY-MM-DDTHH-MM-SS__episode{episode}__{title}.pkl
+ file_save_name: will be formatted path_folder\YYYY-MM-DDTHH-MM-SS__{title}__EPISODE_{episode}_DATA_STORAGE.pkl
 
 
  :param path_folder: Path to folder
@@ -232,7 +232,7 @@ def set_up_episode_storage(self, path_folder: str, vehicle: AUVSim, step_size: f
  utc_str = datetime.datetime.utcnow().strftime('%Y_%m_%dT%H_%M_%S')
  if len(path_folder) > 0:
  os.makedirs(path_folder, exist_ok=True) # Create folder if not exists yet
- self.file_save_name = os.path.join(path_folder, f"{utc_str}__episode{episode}__{title}.pkl")
+ self.file_save_name = os.path.join(path_folder, f"{utc_str}__{title}__EPISODE_{episode}_DATA_STORAGE.pkl")
  self.vehicle = vehicle # Vehicle instance (not a copy, automatically a reference which is updated in reference)
  if shapes is None:
  shapes = []

main.py

@@ -1,3 +1,4 @@
+import matplotlib.pyplot as plt
 import numpy as np
 import gym
 import gym_dockauv
@@ -11,7 +12,7 @@
 import gym_dockauv.train as train
 
 if __name__ == "__main__":
- train.train(total_timesteps=20000)
+ train.train(total_timesteps=10000)
  train.predict()
  train.post_analysis_directory()
 
@@ -53,3 +54,53 @@
 # env.reset()
 # done = False
 
+
+# import gym_dockauv.objects.shape as shape
+# import time
+#
+#
+# def test_not_vec(l1, ld, cap1, cap2, cap_rad=0.1):
+# for r1, rd in zip(l1, ld):
+# dist = shape.intersec_dist_line_capsule(l1=r1, ld=rd, cap1=cap1, cap2=cap2, cap_rad=cap_rad)
+#
+#
+# def test_vectorized(l1, ld, cap1, cap2, cap_rad=0.1):
+# dist = shape.intersec_dist_line_capsule_vectorized(l1=l1, ld=ld, cap1=cap1, cap2=cap2, cap_rad=cap_rad)
+
+
+# if __name__ == '__main__':
+# # Init
+# t_not_vec = []
+# t_vec = []
+# jj = 100 # Number of function executions in a run
+# # # Loop over different amount of rays
+# for i in range(1, 100):
+# # Create arrays
+# l1 = np.random.random((i, 3))
+# ld = np.random.random((i, 3))
+# cap1 = np.random.random(3)
+# cap2 = np.random.random(3)
+#
+# # Find the sum over xx calls
+# loop = [[], []] # Index 0, not vec, Index1, vec
+# for j in range(0, jj):
+# t = time.process_time()
+# test_not_vec(l1, ld, cap1, cap2)
+# elapsed_time = time.process_time() - t
+# loop[0].append(elapsed_time)
+#
+# t = time.process_time()
+# test_vectorized(l1, ld, cap1, cap2)
+# elapsed_time = time.process_time() - t
+# loop[1].append(elapsed_time)
+#
+# t_not_vec.append(sum(loop[0]))
+# t_vec.append(sum(loop[1]))
+#
+# plt.figure()
+# plt.plot(t_not_vec, label="Not vectorized")
+# plt.plot(t_vec, label="Vectorized")
+# plt.legend()
+# plt.xlabel("Number of rays")
+# plt.ylabel(f"t [s] over {jj} runs")
+# plt.show()