gym-examples 3.0.266__py3-none-any.whl → 3.0.268__py3-none-any.whl

gym_examples/__init__.py

@@ -5,4 +5,4 @@ register(
      entry_point="gym_examples.envs:WSNRoutingEnv",
 )
 
-__version__ = "3.0.266"
+__version__ = "3.0.268"

gym_examples/envs/wsn_env.py

@@ -25,7 +25,7 @@ initial_number_of_packets = 1  # initial number of packets to transmit
 latency_per_hop = 1  # latency per hop in seconds
 
 base_back_up_dir = "results/data/"
-max_reward = 5 # maximum reward value when the sensors sent data to the base station. The opposite value is when the sensors perform an unauthorized action
+max_reward = 3 # maximum reward value when the sensors sent data to the base station. The opposite value is when the sensors perform an unauthorized action
 
 # Define the final reward function using an attention mechanism
 class Attention(nn.Module):
@@ -73,7 +73,7 @@ class WSNRoutingEnv(gym.Env):
         self.episode_count = 0
         self.scale_displacement = 0.01 * (upper_bound - lower_bound) # scale of the random displacement of the sensors
         self.epsilon = 1e-10 # small value to avoid division by zero
-        # Initialize the position of the sensors randomly
+        self.current_sensor = 0 # Index of the current sensor
 
         # Define observation space
         self.observation_space = Tuple(
@@ -81,28 +81,13 @@ class WSNRoutingEnv(gym.Env):
         )
 
         # self.action_space = Tuple(tuple([Discrete(self.n_sensors + 1)] * self.n_agents))
-        self.action_space = MultiDiscrete([self.n_sensors + 1] * self.n_agents)
+        # self.action_space = MultiDiscrete([self.n_sensors + 1] * self.n_agents)
         # self.action_space = MultiDiscrete([self.n_agents, self.n_sensors + 1]) 
-        # self.action_space = Discrete(self.n_agents * (self.n_sensors + 1))
+        self.action_space = Discrete(self.n_sensors + 1)  # +1 for the base station
 
         self.reset()        
 
     def reset(self):
-        
-        # if self.episode_count > 1 and os.getenv('PRINT_STATS') == 'True': # Statistics for the PPO algorithm during the training phase
-        #     self.episode_returns.append(self.episode_return)
-        #     self.episode_std_remaining_energy.append(np.std(self.remaining_energy))
-        #     self.episode_mean_remaining_energy.append(np.mean(self.remaining_energy))
-        #     self.episode_total_consumption_energy.append(np.sum(initial_energy - self.remaining_energy))
-        #     self.episode_network_throughput.append(self.network_throughput)
-        #     self.episode_packet_delivery_ratio.append(self.packet_delivery_ratio)
-        #     self.episode_network_lifetime.append(self.network_lifetime)
-        #     self.episode_average_latency.append(self.average_latency)
-        
-        # if self.episode_count > 1 and os.getenv('PRINT_STATS') == 'True-False': # Statistics for the PPO algorithm during the evaluation phase
-        #     print(f"Episode: {self.episode_count}")           
-        #     print(self.get_metrics())  
-
         self.episode_return = 0              
         self.sensor_positions = np.random.rand(self.n_sensors, 2) * (upper_bound - lower_bound) + lower_bound
         self.distance_to_base = np.linalg.norm(self.sensor_positions - base_station_position, axis=1)
@@ -131,18 +116,13 @@ class WSNRoutingEnv(gym.Env):
 
 
     def step(self, actions):
-        actions = [actions[i] for i in range(self.n_agents)] # We want to go back from the MultiDiscrete action space to a tuple of tuple of Discrete action spaces
         self.steps += 1 
-        # rewards = [-max_reward] * self.n_sensors
-        rewards = [0] * self.n_sensors
-        dones = [False] * self.n_sensors
-        for i, action in enumerate(actions):
-            if action not in range(self.n_sensors + 1):
-                raise ValueError("Invalid action!")
-            
-            if i >= self.n_sensors:
-                raise ValueError("Invalid sensor index!") # the number of actions is greater than the number of sensors
-
+        rewards = - max_reward
+        # rewards = 0
+        dones = False
+        actions = np.array([self.current_sensor, actions])
+        print(f"\nactions from step in WSNRoutingEnv: {actions}\n")
+        for i, action in [actions]:
             if self.remaining_energy[i] <= 0 or self.number_of_packets[i] <= 0:
                 continue  # Skip if sensor has no energy left or no packets to transmit
             
@@ -157,6 +137,11 @@ class WSNRoutingEnv(gym.Env):
                 transmission_energy = self.transmission_energy(self.number_of_packets[i], self.distance_to_base[i])
                 if self.remaining_energy[i] < transmission_energy:
                     self.remaining_energy[i] = 0
+                    next_sensor = self.find_next_sensor()
+                    if next_sensor is None:
+                        dones = True
+                    else:
+                        self.current_sensor = next_sensor
                     continue  # Skip if the sensor does not have enough energy to transmit data to the base station
                 
                 self.update_sensor_energies(i, transmission_energy)
@@ -168,10 +153,12 @@ class WSNRoutingEnv(gym.Env):
                 self.total_latency += self.packet_latency[i] + latency_per_hop
                 self.packet_latency[i] = 0
 
-                # rewards[i] = self.compute_individual_rewards(i, action)
-                # rewards[i] = np.ones(input_dim) * max_reward # Reward for transmitting data to the base station
-                rewards[i] = np.ones(input_dim) # Reward for transmitting data to the base station
-                dones[i] = True       
+                rewards = max_reward # Reward for transmitting data to the base station
+                next_sensor = self.find_next_sensor()
+                if next_sensor is None:
+                    dones = True
+                else:
+                    self.current_sensor = next_sensor
             else:
                 distance = np.linalg.norm(self.sensor_positions[i] - self.sensor_positions[action])
                 if distance > self.coverage_radius:
@@ -180,11 +167,21 @@ class WSNRoutingEnv(gym.Env):
                 transmission_energy = self.transmission_energy(self.number_of_packets[i], distance)
                 reception_energy = self.reception_energy(self.number_of_packets[i])
                 if self.remaining_energy[i] < transmission_energy:
-                    self.remaining_energy[i] = 0 
+                    self.remaining_energy[i] = 0
+                    next_sensor = self.find_next_sensor()
+                    if next_sensor is None:
+                        dones = True
+                    else:
+                        self.current_sensor = next_sensor 
                     continue  # Skip if the sensor does not have enough energy to transmit data to the next hop
                 if self.remaining_energy[action] < reception_energy:
                     self.number_of_packets[i] = 0
                     self.remaining_energy[action] = 0
+                    next_sensor = self.find_next_sensor()
+                    if next_sensor is None:
+                        dones = True
+                    else:
+                        self.current_sensor = next_sensor
                     continue  # Skip if the next hop does not have enough energy to receive data
 
                 self.update_sensor_energies(i, transmission_energy)  
@@ -196,19 +193,17 @@ class WSNRoutingEnv(gym.Env):
                 self.packet_latency[action] += self.packet_latency[i] + latency_per_hop
                 self.packet_latency[i] = 0
 
-                rewards[i] = self.compute_individual_rewards(i, action)  
+                rewards = self.compute_individual_rewards(i, action)  
 
                 # Update the number of packets
                 self.number_of_packets[action] += self.number_of_packets[i]
+                
+                self.current_sensor = action
             self.number_of_packets[i] = 0 # Reset the number of packets of the sensor i
             # Calculate final reward
             # rewards[i] = self.compute_attention_rewards(rewards[i])
-            rewards[i] = np.mean(rewards[i])
-            # rewards[i] = self.compute_weighted_sum_rewards(rewards[i])
-        for i in range(self.n_sensors):
-            if (self.remaining_energy[i] <= 0) or (self.number_of_packets[i] <= 0):
-                dones[i] = True
-
+            rewards = np.mean(rewards)
+        
         # Integrate the mobility of the sensors
         # self.integrate_mobility() 
 
@@ -219,13 +214,10 @@ class WSNRoutingEnv(gym.Env):
 
         self.get_metrics()
 
-        rewards = [r.item() if isinstance(r, torch.Tensor) else r for r in rewards] # Convert the rewards to a list of floats
-        # rewards = np.sum(rewards)  # Sum the rewards of all agents
-        rewards = np.mean(rewards)  # Average the rewards of all agents
-        # rewards = np.mean(self.compute_network_rewards())  # Average the rewards of all agents
-        # print(f"Step: {self.steps}, Rewards: {rewards}, Done: {dones}")
-        dones = all(dones)  # Done if all agents are done
-        
+        rewards = rewards.item() if isinstance(rewards, torch.Tensor) else rewards # Convert the reward to a float
+        if not dones:
+            dones = all(self.remaining_energy[i] <= 0 or self.number_of_packets[i] == 0 for i in range(self.n_sensors))
+
         return self._get_obs(), rewards, dones, self.get_metrics()
 
 
@@ -233,7 +225,8 @@ class WSNRoutingEnv(gym.Env):
         return [{'remaining_energy': np.array([e]), 
                  'consumption_energy': np.array([initial_energy - e]),
                  'sensor_positions': p,
-                 'number_of_packets': np.array([d])
+                 'number_of_packets': np.array([d]),
+                 'curent_sensor': np.array([self.current_sensor])
                 } for e, p, d in zip(self.remaining_energy, self.sensor_positions, self.number_of_packets)]
 
 
@@ -242,7 +235,8 @@ class WSNRoutingEnv(gym.Env):
             'remaining_energy': Box(low=0, high=initial_energy, shape=(1,), dtype=np.float64),
             'consumption_energy': Box(low=0, high=initial_energy, shape=(1,), dtype=np.float64),
             'sensor_positions': Box(low=lower_bound, high=upper_bound, shape=(2,), dtype=np.float64),
-            'number_of_packets': Box(low=0, high=self.n_sensors * initial_number_of_packets + 1, shape=(1,), dtype=int)
+            'number_of_packets': Box(low=0, high=self.n_sensors * initial_number_of_packets + 1, shape=(1,), dtype=int),
+            'current_sensor': Box(low=0, high=self.n_sensors - 1, shape=(1,), dtype=int)
         })
 
 
@@ -496,4 +490,12 @@ class WSNRoutingEnv(gym.Env):
             "packet_delivery_ratio": self.packet_delivery_ratio,
             "network_lifetime": self.network_lifetime,
             "average_latency": self.average_latency
-        }
+        }
+    
+
+    def find_next_sensor(self):
+        for offset in range(1, self.n_sensors):
+            next_index = (self.current_sensor + offset) % self.n_sensors
+            if self.remaining_energy[next_index] > 0 and self.number_of_packets[next_index] > 0:
+                return next_index
+        return None  # If no such sensor is found

{gym_examples-3.0.266.dist-info → gym_examples-3.0.268.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: gym-examples
-Version: 3.0.266
+Version: 3.0.268
 Summary: A custom environment for multi-agent reinforcement learning focused on WSN routing.
 Home-page: https://github.com/gedji/CODES.git
 Author: Georges Djimefo

gym_examples-3.0.268.dist-info/RECORD

@@ -0,0 +1,7 @@
+gym_examples/__init__.py,sha256=HpffWs56Jd1UDNQGyBLzk4OLzrDNZMUx_SN58BZaJA8,166
+gym_examples/envs/__init__.py,sha256=lgMe4pyOuUTgTBUddM0iwMlETsYTwFShny6ifm8PGM8,53
+gym_examples/envs/wsn_env.py,sha256=MTP_6sqJT4T2IBGedoLOGaKVZras8gT2NXvWHfggOXA,24154
+gym_examples-3.0.268.dist-info/METADATA,sha256=NvjGw3Ag9m0cr-a5mu_DQn2IkoEiiCUmZjeFAQye0wI,412
+gym_examples-3.0.268.dist-info/WHEEL,sha256=2wepM1nk4DS4eFpYrW1TTqPcoGNfHhhO_i5m4cOimbo,92
+gym_examples-3.0.268.dist-info/top_level.txt,sha256=rJRksoAF32M6lTLBEwYzRdo4PgtejceaNnnZ3HeY_Rk,13
+gym_examples-3.0.268.dist-info/RECORD,,

gym_examples-3.0.266.dist-info/RECORD

@@ -1,7 +0,0 @@
-gym_examples/__init__.py,sha256=KqEqczXOEfVmbkVd7dsDXk4gyzc5VC_HrycnPdSH_8M,166
-gym_examples/envs/__init__.py,sha256=lgMe4pyOuUTgTBUddM0iwMlETsYTwFShny6ifm8PGM8,53
-gym_examples/envs/wsn_env.py,sha256=4CXrSu-8epidbKW7JgPIzbXPS-WPtCC82FIsdPPAlMQ,24863
-gym_examples-3.0.266.dist-info/METADATA,sha256=KdqFD13218m6mSDcVJ5fpru17gAS0IuYts1m4uMPUTE,412
-gym_examples-3.0.266.dist-info/WHEEL,sha256=2wepM1nk4DS4eFpYrW1TTqPcoGNfHhhO_i5m4cOimbo,92
-gym_examples-3.0.266.dist-info/top_level.txt,sha256=rJRksoAF32M6lTLBEwYzRdo4PgtejceaNnnZ3HeY_Rk,13
-gym_examples-3.0.266.dist-info/RECORD,,