ball position reward

citbrains · Mar 7, 2024 · 35a7e5f · 35a7e5f
1 parent 6a61dd7
commit 35a7e5f
Show file tree

Hide file tree

Showing 2 changed files with 1 addition and 2 deletions.
diff --git a/controllers/mat_train/runner/soccer_runner.py b/controllers/mat_train/runner/soccer_runner.py
@@ -101,7 +101,6 @@ def run(self):
                 self.log_train(train_infos, total_num_steps)
 
                 num = self.episode_length * self.log_interval
-                print("rewards: {}".format(sum_episode_rewards[0]/num))
                 self.writter.add_scalars("train_episode_rewards", {"total_rewards": sum_episode_rewards[0]/num}, total_num_steps)
                 sum_episode_rewards = [0 for _ in range(self.n_rollout_threads)]
                 self.writter.add_scalars("train_episode_rewards", {"ball_distance_reward": sum_ball_distance_reward[0]/self.num_agents/num}, total_num_steps)

diff --git a/controllers/mat_train/soccer/soccer.py b/controllers/mat_train/soccer/soccer.py
@@ -188,7 +188,7 @@ def step(self, action):
             velocity_reward = 10.0
             out_of_field_reward = -100.0
             collision_reward = -1.0
-            ball_position_reward = 2.0
+            ball_position_reward = 4.0
             ball_tracking_reward = 0.1
 
             rew_ball_distance = dict(zip(self.agents, [0.0 for _ in self.agents]))