Update app.py

app.py

@@ -4,8 +4,8 @@ import requests
 from geopy.geocoders import Nominatim  # Valid import
 from stable_baselines3 import PPO
 from stable_baselines3.common.vec_env import DummyVecEnv
-import gym
-from gym import spaces
+import gymnasium as gym
+from gymnasium import spaces
 import numpy as np
 from PIL import Image
 import gradio as gr
@@ -68,7 +68,11 @@ class TrafficSimEnv(gym.Env):
         self.congestion_level = congestion_level
 
         # Define observation space: [congestion_level, signal_duration]
-        self.observation_space = spaces.Box(low=np.array([0, 0]), high=np.array([10, 60]), dtype=np.float32)
+        self.observation_space = spaces.Box(
+            low=np.array([0, 0], dtype=np.float32),
+            high=np.array([10, 60], dtype=np.float32),
+            dtype=np.float32
+        )
 
         # Define action space: [increase, decrease, maintain]
         self.action_space = spaces.Discrete(3)
@@ -129,14 +133,8 @@ class TrafficSimEnv(gym.Env):
 def optimize_signal_rl(congestion_level):
     try:
         # Map congestion levels (string to numeric)
-        congestion_map = {
-            "Low": 2,
-            "Medium": 5,
-            "High": 8
-        }
-
-        if isinstance(congestion_level, str):
-            congestion_level = congestion_map.get(congestion_level, 5)  # Default to "Medium" if unrecognized
+        congestion_map = {"Low": 2, "Medium": 5, "High": 8}
+        congestion_level = congestion_map.get(congestion_level, 5) if isinstance(congestion_level, str) else congestion_level
 
         # Create environment
         env = DummyVecEnv([lambda: TrafficSimEnv(congestion_level)])
@@ -149,27 +147,29 @@ def optimize_signal_rl(congestion_level):
 
         # Reset the environment
         obs = env.reset()
+        logging.debug(f"Observation after reset: {obs}, type: {type(obs)}, shape: {np.shape(obs)}")
 
-        # Ensure obs is properly handled (flatten if necessary)
-        if obs.ndim == 0:
+        # Ensure obs is properly handled
+        if isinstance(obs, np.ndarray) and obs.ndim == 0:
             obs = np.expand_dims(obs, axis=0)  # Convert 0D array to 1D
+        elif not isinstance(obs, np.ndarray):
+            obs = np.array(obs)  # Ensure it's an array
 
         # RL Optimization loop
         for _ in range(10):
             action, _ = model.predict(obs, deterministic=True)
             obs, rewards, dones, infos = env.step(action)
+            logging.debug(f"Step results -> obs: {obs}, rewards: {rewards}, dones: {dones}, infos: {infos}")
 
             # Ensure obs is properly handled
-            if obs.ndim == 0:
+            if isinstance(obs, np.ndarray) and obs.ndim == 0:
                 obs = np.expand_dims(obs, axis=0)
+            elif not isinstance(obs, np.ndarray):
+                obs = np.array(obs)
 
-            # Extract single observation and reward
-            obs = obs[0]
-            reward = rewards[0]
-            done = dones[0]
-
-            # Debugging logs for better tracking
-            logging.debug(f"Action: {action}, Observation: {obs}, Reward: {reward}, Done: {done}")
+            obs = obs[0] if obs.ndim > 1 else obs  # Flatten to 1D if necessary
+            reward = rewards[0] if isinstance(rewards, list) else rewards
+            done = dones[0] if isinstance(dones, list) else dones
 
             if done:
                 break