Add files via upload

env/envs.py

@@ -1,7 +1,7 @@
 import gym
 from gym.spaces import Box, Dict
 import numpy as np
-# from sl_pso import SLPSO as meta_optimizer
+# from utils.metaheuristics import SLPSO as PSO
 
 
 class LazyAgentsCentralized(gym.Env):
@@ -23,6 +23,7 @@ class LazyAgentsCentralized(gym.Env):
  # TODO (9): [o] Implement angle WRAPPING <- not necessary as we use the relative angle and sin/cos(θ)
  # just added the function to the env class (wrapped_ang = self.wrap_to_pi(ang))
  # TODO (10): [o] Configuration template
+ # TODO (11): [o] Allow auto_step as well as the norminal step, one_step
 
  def __init__(self, config):
  """
@@ -52,6 +53,9 @@ def __init__(self, config):
  "std_pos_rate_converged": 0.1, # Standard position rate when converged. Default is 0.1
  "std_vel_rate_converged": 0.2, # Standard velocity rate when converged. Default is 0.2
  "max_time_step": 1000 # Maximum time steps. Default is 1000,
+
+ # Step mode
+ "auto_step": False # If True, the env will step automatically (i.e. episode length==1). Default is False
  }
  """
 
@@ -89,6 +93,7 @@ def __init__(self, config):
  self.std_vel_rate_converged = self.config["std_vel_rate_converged"] \
  if "std_vel_rate_converged" in self.config else 0.2 # m/s
  self.max_time_step = self.config["max_time_step"] if "max_time_step" in self.config else 1000
+ self.do_auto_step = self.config["auto_step"] if "auto_step" in self.config else False
 
  # Define action space
  # Laziness vector; padding included
@@ -346,6 +351,32 @@ def view_adjacency_matrix(self, mask=None):
  def step(self,
  action: np.ndarray,
  ):
+ if self.do_auto_step:
+ obs, reward, done, info = self.auto_step(action)
+ else:
+ obs, reward, done, info = self.single_step(action)
+
+ return obs, reward, done, info
+
+ def auto_step(self,
+ action: np.ndarray,
+ ):
+ # The given action is continuously used in all single_step calls until the episode is done.
+
+ obs = self.get_observation()
+ episode_reward = 0
+ done = False
+ info = {}
+ constant_action = action
+ while not done:
+ obs, reward, done, info = self.single_step(constant_action)
+ episode_reward += reward
+
+ return obs, episode_reward, done, info
+
+ def single_step(self,
+ action: np.ndarray,
+ ):
  # Note: action is not used as control input;
  # it is used to weight the control input (i.e. laziness)