some tiny change

scripts/gyms/Kick_3M.py

@@ -256,7 +256,7 @@ class Kick_3M(gym.Env):
                 waiting_steps += 1
             dis = np.linalg.norm(self.ball_pos - w.ball_cheat_abs_pos)
 
-            reward = 10 - abs(3 - dis) - abs(w.ball_cheat_abs_pos[1] - self.ball_pos[1])
+            reward = 3 - abs(3 - dis) - abs(w.ball_cheat_abs_pos[1] - self.ball_pos[1])
             if self.isfallen or w.ball_cheat_abs_pos[0] < self.ball_pos[0]:
                 reward = 0
             # print(reward)
@@ -282,7 +282,7 @@ class Train(Train_Base):
         minibatch_size = 64  # should be a factor of (n_steps_per_env * n_envs)
         total_steps = 30000000
         learning_rate = 3e-5
-        folder_name = f'kick_moving_ball_R{self.robot_type}'
+        folder_name = f'Kick_3M_R{self.robot_type}'
         model_path = f'./scripts/gyms/logs/{folder_name}/'
 
         # print("Model path:", model_path)

train.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+source /home/apollo/下载/Anaconda/etc/profile.d/conda.sh
+conda activate gym
+
+export MKL_SERVICE_FORCE_INTEL=1
+export MKL_THREADING_LAYER=GNU
+export LD_LIBRARY_PATH="/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH"
+
+pkill -f rcssserver3d 2>/dev/null || true
+pkill -f simspark 2>/dev/null || true
+
+sleep 2
+
+python Run_Utils.py