added rendering to DMC envs and updated examples

alr_envs/examples/examples_dmc.py

@@ -6,19 +6,24 @@ def example_dmc(env_name="fish-swim", seed=1, iterations=1000):
     env = make_env(env_name, seed)
     rewards = 0
     obs = env.reset()
-    print(obs)
+    print("observation shape:", env.observation_space.shape)
+    print("action shape:", env.action_space.shape)
 
     # number of samples(multiple environment steps)
-    for i in range(10):
+    for i in range(iterations):
         ac = env.action_space.sample()
         obs, reward, done, info = env.step(ac)
         rewards += reward
 
+        env.render("human")
+
         if done:
-            print(rewards)
+            print(env_name, rewards)
             rewards = 0
             obs = env.reset()
 
+    env.close()
+
 
 def example_custom_dmc_and_mp(seed=1):
     """
@@ -50,12 +55,13 @@ def example_custom_dmc_and_mp(seed=1):
         "weights_scale": 50,
         "goal_scale": 0.1
     }
-    env = make_dmp_env(base_env, wrappers=wrappers, seed=seed, **mp_kwargs)
+    env = make_dmp_env(base_env, wrappers=wrappers, seed=seed, mp_kwargs=mp_kwargs)
     # OR for a deterministic ProMP:
     # env = make_detpmp_env(base_env, wrappers=wrappers, seed=seed, **mp_args)
 
     rewards = 0
     obs = env.reset()
+    env.render("human")
 
     # number of samples/full trajectories (multiple environment steps)
     for i in range(10):
@@ -64,17 +70,26 @@ def example_custom_dmc_and_mp(seed=1):
         rewards += reward
 
         if done:
-            print(rewards)
+            print(base_env, rewards)
             rewards = 0
             obs = env.reset()
 
+    env.close()
+
 
 if __name__ == '__main__':
     # Disclaimer: DMC environments require the seed to be specified in the beginning.
     # Adjusting it afterwards with env.seed() is not recommended as it does not affect the underlying physics.
 
-    # Standard DMC task
-    example_dmc("fish_swim", seed=10, iterations=1000)
+    # For rendering DMC
+    # export MUJOCO_GL="osmesa"
+
+    # Standard DMC Suite tasks
+    example_dmc("fish-swim", seed=10, iterations=100)
+
+    # Manipulation tasks
+    # The vision versions are currently not integrated
+    example_dmc("manipulation-reach_site_features", seed=10, iterations=100)
 
     # Gym + DMC hybrid task provided in the MP framework
     example_dmc("dmc_ball_in_cup_dmp-v0", seed=10, iterations=10)

alr_envs/examples/examples_general.py

@@ -8,7 +8,7 @@ from alr_envs.utils.make_env_helpers import make_env
 from alr_envs.utils.mp_env_async_sampler import AlrContextualMpEnvSampler, AlrMpEnvSampler, DummyDist
 
 
-def example_general(env_id='alr_envs:ALRReacher-v0', seed=1):
+def example_general(env_id: str, seed=1, iterations=1000):
     """
     Example for running any env in the step based setting.
     This also includes DMC environments when leveraging our custom make_env function.
@@ -17,16 +17,16 @@ def example_general(env_id='alr_envs:ALRReacher-v0', seed=1):
     env = make_env(env_id, seed)
     rewards = 0
     obs = env.reset()
-    print("Observation shape: ", obs.shape)
+    print("Observation shape: ", env.observation_space.shape)
     print("Action shape: ", env.action_space.shape)
 
     # number of environment steps
-    for i in range(10000):
+    for i in range(iterations):
         obs, reward, done, info = env.step(env.action_space.sample())
         rewards += reward
 
-        # if i % 1 == 0:
-        #     env.render()
+        if i % 1 == 0:
+            env.render()
 
         if done:
             print(rewards)
@@ -65,10 +65,5 @@ def example_async(env_id="alr_envs:HoleReacherDMP-v0", n_cpu=4, seed=int('533D',
 
 
 if __name__ == '__main__':
-    # DMC
-    # example_general("fish-swim")
-
-    # custom mujoco env
-    # example_general("alr_envs:ALRReacher-v0")
-
-    example_general("ball_in_cup-catch")
+    # Mujoco task from framework
+    example_general("alr_envs:ALRReacher-v0")

alr_envs/examples/examples_motion_primitives.py

@@ -83,12 +83,17 @@ def example_custom_mp(seed=1):
         "weights_scale": 50,
         "goal_scale": 0.1
     }
-    env = make_dmp_env(base_env, wrappers=wrappers, seed=seed, **mp_kwargs)
+    env = make_dmp_env(base_env, wrappers=wrappers, seed=seed, mp_kwargs=mp_kwargs)
     # OR for a deterministic ProMP:
     # env = make_detpmp_env(base_env, wrappers=wrappers, seed=seed)
 
     rewards = 0
-    # env.render(mode=None)
+    # render full DMP trajectory
+    # It is only required to call render() once in the beginning, which renders every consecutive trajectory.
+    # Resetting to no rendering, can be achieved by render(mode=None).
+    # It is also possible to change them mode multiple times when
+    # e.g. only every nth trajectory should be displayed.
+    env.render(mode="human")
     obs = env.reset()
 
     # number of samples/full trajectories (multiple environment steps)
@@ -97,12 +102,6 @@ def example_custom_mp(seed=1):
         obs, reward, done, info = env.step(ac)
         rewards += reward
 
-        if i % 1 == 0:
-            # render full DMP trajectory
-            # render can only be called once in the beginning as well. That would render every trajectory
-            # Calling it after every trajectory allows to modify the mode. mode=None, disables rendering.
-            env.render(mode="human")
-
         if done:
             print(rewards)
             rewards = 0