Add interface for envs controlable by a PD Controller and add more infos to mp_wrapper info return value

alr_envs/mujoco/alr_mujoco_env.py

@@ -7,7 +7,9 @@ from gym import error, spaces
 from gym.utils import seeding
 import numpy as np
 from os import path
-import gym
+
+from alr_envs.utils.mps.alr_env import AlrEnv
+from alr_envs.utils.positional_env import PositionalEnv
 
 try:
     import mujoco_py
@@ -33,7 +35,7 @@ def convert_observation_to_space(observation):
     return space
 
 
-class AlrMujocoEnv(gym.Env):
+class AlrMujocoEnv(PositionalEnv, AlrEnv):
     """
     Superclass for all MuJoCo environments.
     """
@@ -44,7 +46,7 @@ class AlrMujocoEnv(gym.Env):
         Args:
             model_path: path to xml file
             n_substeps: how many steps mujoco does per call to env.step
-            use_servo: use actuator defined in xml, use False for direct torque control
+            apply_gravity_comp: Whether gravity compensation should be active
         """
         if model_path.startswith("/"):
             fullpath = model_path
@@ -73,10 +75,6 @@ class AlrMujocoEnv(gym.Env):
 
         self._set_action_space()
 
-        # action = self.action_space.sample()
-        # observation, _reward, done, _info = self.step(action)
-        # assert not done
-
         observation = self._get_obs()  # TODO: is calling get_obs enough? should we call reset, or even step?
 
         self._set_observation_space(observation)
@@ -204,7 +202,7 @@ class AlrMujocoEnv(gym.Env):
 
         try:
             self.sim.step()
-        except mujoco_py.builder.MujocoException as e:
+        except mujoco_py.builder.MujocoException:
             error_in_sim = True
 
         return error_in_sim

alr_envs/mujoco/ball_in_a_cup/ball_in_a_cup.py

@@ -16,8 +16,6 @@ class ALRBallInACupEnv(alr_mujoco_env.AlrMujocoEnv, utils.EzPickle):
         self._q_vel = []
         # self.weight_matrix_scale = 50
         self.max_ctrl = np.array([150., 125., 40., 60., 5., 5., 2.])
-        self.p_gains = 1 / self.max_ctrl * np.array([200, 300, 100, 100, 10, 10, 2.5])
-        self.d_gains = 1 / self.max_ctrl * np.array([7, 15, 5, 2.5, 0.3, 0.3, 0.05])
 
         self.j_min = np.array([-2.6, -1.985, -2.8, -0.9, -4.55, -1.5707, -2.7])
         self.j_max = np.array([2.6, 1.985, 2.8, 3.14159, 1.25, 1.5707, 2.7])