From 7c9a687770c184ac2b524d9f0c136e0c3b9eed97 Mon Sep 17 00:00:00 2001
From: god <2501229495@qq.com>
Date: Mon, 15 Dec 2025 22:33:08 +0800
Subject: [PATCH] =?UTF-8?q?=E6=96=B0=E5=A2=9E=20DoubleDQN=20=E8=AE=AD?=
 =?UTF-8?q?=E7=BB=83=E4=B8=8E=E8=AF=84=E4=BC=B0=E8=84=9A=E6=9C=AC?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 src/Car-Racing/configs/double_dqn.yaml        |  7 +++
 .../doubledqn_model/eval_doubledqn.py         | 45 +++++++++++++++++++
 src/Car-Racing/{ => dqn_model}/DQN_model.py   |  0
 src/Car-Racing/{ => dqn_model}/eval_dqn.py    |  0
 .../{ => dqn_model}/training_dqn.py           |  0
 5 files changed, 52 insertions(+)
 create mode 100644 src/Car-Racing/configs/double_dqn.yaml
 create mode 100644 src/Car-Racing/doubledqn_model/eval_doubledqn.py
 rename src/Car-Racing/{ => dqn_model}/DQN_model.py (100%)
 rename src/Car-Racing/{ => dqn_model}/eval_dqn.py (100%)
 rename src/Car-Racing/{ => dqn_model}/training_dqn.py (100%)

diff --git a/src/Car-Racing/configs/double_dqn.yaml b/src/Car-Racing/configs/double_dqn.yaml
new file mode 100644
index 0000000000..b573651ac8
--- /dev/null
+++ b/src/Car-Racing/configs/double_dqn.yaml
@@ -0,0 +1,7 @@
+# Extends base DQN config
+_extends: ./dqn.yaml
+
+hyperparameters:
+  # Double DQN specific
+  tau: 0.005                    # For soft target updates
+  update_target_every: 10000     # Less frequent hard updates
\ No newline at end of file
diff --git a/src/Car-Racing/doubledqn_model/eval_doubledqn.py b/src/Car-Racing/doubledqn_model/eval_doubledqn.py
new file mode 100644
index 0000000000..bdc3c02dd6
--- /dev/null
+++ b/src/Car-Racing/doubledqn_model/eval_doubledqn.py
@@ -0,0 +1,45 @@
+import os
+import sys
+sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))
+
+import double_dqn_model.double_dqn as DDQN
+import gymnasium as gym
+from gymnasium.wrappers import GrayscaleObservation, ResizeObservation, FrameStackObservation, RecordVideo
+
+# Load the saved model
+save_dir = 'training\saved_models'
+model_file = 'DoubleDQN.pt'
+driver = DDQN.DoubleDQNAgent(
+    state_space_shape=(4, 84, 84),
+    action_n=5,
+    load_state=True,
+    load_model=model_file
+)
+
+# Evaluate
+def evaluate_agent(agent, num_episodes=5, render=True):
+    env = gym.make("CarRacing-v3", continuous=False, render_mode="rgb_array")
+    env = RecordVideo(env, video_folder='videos\DoubleDQN')
+    env = DDQN.SkipFrame(env, skip=4)
+    env = GrayscaleObservation(env)
+    env = ResizeObservation(env, (84, 84))
+    env = FrameStackObservation(env, stack_size=4)
+    agent.epsilon = 0
+    seeds_list = [i for i in range(num_episodes)]
+    scores = []
+    for episode, seed in enumerate(seeds_list):
+        state, info = env.reset(seed=seed)
+        score = 0
+        updating = True
+        while updating:
+            action = agent.take_action(state)
+            state, reward, terminated, truncated, info = env.step(action)
+            score += reward
+            updating = not (terminated or truncated)
+        scores.append(score)
+        print(f"Evaluation Episode {episode+1}/{num_episodes} | Seed: {seed} | Score: {score:.1f}")
+    env.close()
+    return sum(scores) / len(scores)
+
+avg_score = evaluate_agent(driver, num_episodes=5, render=True)
+print(f"Average evaluation score: {avg_score:.1f}")
\ No newline at end of file
diff --git a/src/Car-Racing/DQN_model.py b/src/Car-Racing/dqn_model/DQN_model.py
similarity index 100%
rename from src/Car-Racing/DQN_model.py
rename to src/Car-Racing/dqn_model/DQN_model.py
diff --git a/src/Car-Racing/eval_dqn.py b/src/Car-Racing/dqn_model/eval_dqn.py
similarity index 100%
rename from src/Car-Racing/eval_dqn.py
rename to src/Car-Racing/dqn_model/eval_dqn.py
diff --git a/src/Car-Racing/training_dqn.py b/src/Car-Racing/dqn_model/training_dqn.py
similarity index 100%
rename from src/Car-Racing/training_dqn.py
rename to src/Car-Racing/dqn_model/training_dqn.py