small bp and tt updates

alr_envs/alr/mujoco/beerpong/beerpong.py

@@ -127,24 +127,28 @@ class ALRBeerBongEnv(MujocoEnv, utils.EzPickle):
             self._steps += 1
         else:
             reward = -30
+            reward_infos = dict()
             success = False
             is_collided = False
             done = True
             ball_pos = np.zeros(3)
             ball_vel = np.zeros(3)
 
-        return ob, reward, done, dict(reward_dist=reward_dist,
-                                      reward_ctrl=reward_ctrl,
-                                      reward=reward,
-                                      velocity=angular_vel,
-                                      # traj=self._q_pos,
-                                      action=a,
-                                      q_pos=self.sim.data.qpos[0:7].ravel().copy(),
-                                      q_vel=self.sim.data.qvel[0:7].ravel().copy(),
-                                      ball_pos=ball_pos,
-                                      ball_vel=ball_vel,
-                                      success=success,
-                                      is_collided=is_collided, sim_crash=crash)
+        infos = dict(reward_dist=reward_dist,
+                     reward_ctrl=reward_ctrl,
+                     reward=reward,
+                     velocity=angular_vel,
+                     # traj=self._q_pos,
+                     action=a,
+                     q_pos=self.sim.data.qpos[0:7].ravel().copy(),
+                     q_vel=self.sim.data.qvel[0:7].ravel().copy(),
+                     ball_pos=ball_pos,
+                     ball_vel=ball_vel,
+                     success=success,
+                     is_collided=is_collided, sim_crash=crash)
+        infos.update(reward_infos)
+
+        return ob, reward, done, infos
 
     def check_traj_in_joint_limits(self):
         return any(self.current_pos > self.j_max) or any(self.current_pos < self.j_min)

alr_envs/alr/mujoco/beerpong/beerpong_reward_staged.py

@@ -110,7 +110,7 @@ class BeerPongReward:
             success = ball_in_cup
             crash = self._is_collided
         else:
-            reward = - 1e-4 * action_cost
+            reward = - 1e-2 * action_cost
             success = False
             crash = False