验证阶段加输出，更新奖励

.gitignore

@@ -9,8 +9,6 @@ __pycache__/
 
 # Pytorch weights
 weights/
-PPO_preTrained/
-PPO_logs/
 logs/
 
 # Distribution / packaging

DDPG_solver/main.py

@@ -125,7 +125,7 @@ def main():
 
                 '''record & log'''
                 if total_steps % opt.eval_interval == 0:
-                    ep_r = evaluate_policy(eval_env, agent, turns=3)
+                    ep_r = evaluate_policy(eval_env, agent, turns=1)
                     if opt.write:
                         writer.add_scalar(
                             'ep_r', ep_r, global_step=total_steps)

env.py

@@ -41,7 +41,7 @@ class PartitionMazeEnv(gym.Env):
         ##############################
         self.CUT_NUM = 4    # 横切一半，竖切一半
         self.BASE_LINE = 4000     # 基准时间，通过greedy或者蒙特卡洛计算出来
-        self.MAX_STEPS = 200        # 迷宫走法步数上限
+        self.MAX_STEPS = 50        # 迷宫走法步数上限
 
         self.phase = 0    # 阶段控制，0：区域划分阶段，1：迷宫初始化阶段，2：走迷宫阶段
         self.partition_step = 0      # 区域划分阶段步数，范围 0~4
@@ -159,7 +159,7 @@ class PartitionMazeEnv(gym.Env):
                 else:
                     # 进入阶段 1：初始化迷宫
                     self.phase = 1
-                    reward = 10
+                    reward = 100
                     
                     # 构建反向索引，方便后续计算
                     self.reverse_rectangles = {v['center']: k for k, v in self.rectangles.items()}
@@ -194,6 +194,8 @@ class PartitionMazeEnv(gym.Env):
             # 查表，找出当前车辆所在的网格
             current_row, current_col = self.reverse_rectangles[self.car_pos[current_car]]
             
+            reward = 0
+
             # 当前动作 a 为 1 维连续动作，映射到四个方向
             if a < 0.2:
                 move_dir = 'up'
@@ -209,16 +211,31 @@ class PartitionMazeEnv(gym.Env):
             # 初始化新的行、列为当前值
             new_row, new_col = current_row, current_col
 
-            if move_dir == 'up' and current_row > 0:
+            if move_dir == 'up':
+                if current_row > 0:
                     new_row = current_row - 1
-            elif move_dir == 'down' and current_row < len(self.row_cuts) - 2:
+                else: # 错误的移动给一些惩罚？
+                    new_row = current_row
+                    # reward -= 10
+            elif move_dir == 'down':
+                if current_row < len(self.row_cuts) - 2:
                     new_row = current_row + 1
-            elif move_dir == 'left' and current_col > 0:
+                else:
+                    new_row = current_row
+                    # reward -= 10
+            elif move_dir == 'left':
+                if current_col > 0:
                     new_col = current_col - 1
-            elif move_dir == 'right' and current_col < len(self.col_cuts) - 2:
+                else:
+                    new_col = current_col
+                    # reward -= 10
+            elif move_dir == 'right':
+                if current_col < len(self.col_cuts) - 2:
                     new_col = current_col + 1
+                else:
+                    new_col = current_col
+                    # reward -= 10
             # 如果移动不合法，或者动作为stay，则保持原位置
-            # TODO 移动不合法，加一些惩罚
 
             # 更新车辆位置
             self.car_pos[current_car] = self.rectangles[(
@@ -235,7 +252,6 @@ class PartitionMazeEnv(gym.Env):
             # 观察状态
             state = np.concatenate(
                 [self.partition_values, np.array(self.car_pos).flatten()])            
-            reward = 0
 
             # Episode 终止条件：所有网格均被访问或步数达到上限
             done = all([value['is_visited'] for _, value in self.rectangles.items()]) or (
@@ -247,7 +263,7 @@ class PartitionMazeEnv(gym.Env):
                 # print(T)
                 # print(self.partition_values)
                 # print(self.car_traj)
-                reward += self.BASE_LINE / T * 100
+                reward += self.BASE_LINE / T * 1000
             elif done and self.step_count >= self.MAX_STEPS:
                 reward += -1000