Merge branch 'master' into four_ages_fix

README.md

@@ -42,6 +42,11 @@ cd loco-mujoco
 pip install -e . 
 ```
 
+> [!NOTE]
+> We fixed the version of MuJoCo to 2.3.7 during installation since we found that there are slight 
+> differences in the simulation, which made testing very difficult. However, in practice, you can 
+> use any newer version of MuJoCo! Just install it after installing LocoMuJoCo.
+
 ### Download the Datasets
 After installing LocoMuJoCo, new commands for downloading the datasets will be setup for you.
 You have the choice of downloading all datasets available or only the ones you need.

docs/conf.py

@@ -8,25 +8,26 @@
 
 import os
 import sys
+
 sys.path.insert(0, os.path.abspath(".."))
 
 
 project = 'LocoMuJoCo'
 copyright = '2024, Firas Al-Hafez'
 author = 'Firas Al-Hafez'
-release = 'v0.2.2'
+release = 'v0.3.0'
 
 # -- General configuration ---------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#general-configuration
 
-extensions = ['sphinx.ext.todo', 'sphinx.ext.viewcode', 'sphinx.ext.autodoc', "m2r2"]
+extensions = ['sphinx.ext.todo', 'sphinx.ext.viewcode', 'sphinx.ext.autodoc', "m2r2", 'sphinx.ext.napoleon']
 
 templates_path = ['_templates']
 exclude_patterns = ['_build', 'Thumbs.db', '.DS_Store']
 add_module_names = False
 autodoc_member_order = 'groupwise'
 source_suffix = [".rst", ".md"]
-
+napoleon_google_docstring = True
 
 # -- Options for HTML output -------------------------------------------------
 # https://www.sphinx-doc.org/en/master/usage/configuration.html#options-for-html-output

docs/source/humanoids/unitreeg1.rst

@@ -0,0 +1,13 @@
+Unitree G1
+=======================================================
+
+.. image:: https://github.com/robfiras/loco-mujoco/assets/69359729/be7c3b29-03db-49c2-8e27-527562218f74
+ :width: 49%
+.. image:: https://github.com/robfiras/loco-mujoco/assets/69359729/23cf5988-7326-453e-af24-0f0adc255c9c
+ :width: 49%
+
+.. automodule:: loco_mujoco.environments.humanoids.unitreeG1
+ :members:
+ :undoc-members:
+ :private-members:
+ :show-inheritance:

docs/source/loco_mujoco.environments.humanoids.rst

@@ -8,6 +8,7 @@ Here you find an overview of the API of all Humanoid environments.
  ./humanoids/talos.rst
  ./humanoids/atlas.rst
  ./humanoids/unitreeh1.rst
+ ./humanoids/unitreeg1.rst
  ./humanoids/torque_humanoid.rst
  ./humanoids/muscle_humanoid.rst
  ./humanoids/base_humanoids.rst

docs/source/loco_mujoco.installation.rst

@@ -17,6 +17,10 @@ or you do an editable installation by cloning this repository and then running:
  cd loco-mujoco
  pip install -e .
 
+.. note::
+ We fixed the version of MuJoCo to 2.3.7 during installation since we found that there are slight
+ differences in the simulation, which made testing very difficult. However, in practice, you can
+ use any newer version of MuJoCo! Just install it after installing LocoMuJoCo.
 
 Download the Datasets
 ---------------------

docs/utils.py

@@ -1,6 +1,6 @@
 import inspect
 import loco_mujoco
-from loco_mujoco.environments import HumanoidTorque, HumanoidMuscle, UnitreeA1
+from loco_mujoco.environments import HumanoidTorque, HumanoidMuscle, UnitreeA1, UnitreeG1
 from mushroom_rl.utils.mujoco import *
 
 
@@ -136,7 +136,10 @@ def get_action_space_table_docs(env, use_muscles=False):
  else:
  motors_to_remove = []
  except :
- action_spec = env._get_action_specification(use_muscles)
+ if type(env) == UnitreeG1:
+ action_spec = env._get_action_specification()
+ else:
+ action_spec = env._get_action_specification(use_muscles)
  _, motors_to_remove, _, _ = env._get_xml_modifications()
 
  header = ["Index", "Name in XML", "Control Min", "Control Max", "Disabled"]
@@ -160,14 +163,14 @@ def get_action_space_table_docs(env, use_muscles=False):
 if __name__ == "__main__":
 
  # # Talos
- env = loco_mujoco.LocoEnv.make("Talos", disable_arms=False, disable_back=False)
- additional_info = [["Mass of the Weight", "0.0", "inf", "Only Enabled for Carry Task", "1", "Mass [kg]"],
- ["3D linear Forces between Right Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"],
- ["3D linear Forces between Left Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"]]
- print(get_obs_space_table_docs(env, additional_info)[0],
- "\nNumber of obs that are on by default: ", get_obs_space_table_docs(env, additional_info)[1], "\n")
- print(get_action_space_table_docs(env)[0],
- "\nNumber of actions that are on by default: ", get_action_space_table_docs(env)[1])
+ # env = loco_mujoco.LocoEnv.make("Talos", disable_arms=False, disable_back=False)
+ # additional_info = [["Mass of the Weight", "0.0", "inf", "Only Enabled for Carry Task", "1", "Mass [kg]"],
+ #  ["3D linear Forces between Right Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"],
+ #  ["3D linear Forces between Left Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"]]
+ # print(get_obs_space_table_docs(env, additional_info)[0],
+ #  "\nNumber of obs that are on by default: ", get_obs_space_table_docs(env, additional_info)[1], "\n")
+ # print(get_action_space_table_docs(env)[0],
+ #  "\nNumber of actions that are on by default: ", get_action_space_table_docs(env)[1])
 
  # # Atlas
  # env = loco_mujoco.LocoEnv.make("Atlas", disable_arms=False, disable_back=False)
@@ -191,6 +194,15 @@ def get_action_space_table_docs(env, use_muscles=False):
  # print(get_action_space_table_docs(env)[0],
  # "\nNumber of actions that are on by default: ", get_action_space_table_docs(env)[1])
 
+ # UnitreeG1
+ env = loco_mujoco.LocoEnv.make("UnitreeG1", disable_arms=False, disable_back=False)
+ additional_info = [["3D linear Forces between Right Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"],
+ ["3D linear Forces between Left Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"]]
+ print(get_obs_space_table_docs(env, additional_info)[0],
+ "\nNumber of obs that are on by default: ", get_obs_space_table_docs(env, additional_info)[1], "\n")
+ print(get_action_space_table_docs(env)[0],
+ "\nNumber of actions that are on by default: ", get_action_space_table_docs(env)[1])
+
  # Torque Humanoid
  # env = loco_mujoco.LocoEnv.make("HumanoidTorque", disable_back=False, disable_arms=False)
  # additional_info = [["3D linear Forces between Back Right Foot and Floor", "0.0", "inf", "True", "3", "Force [N]"],

examples/imitation_learning/confs.yaml

@@ -129,6 +129,25 @@ UnitreeH1:
  use_noisy_targets: False
  last_policy_activation: "identity"
 
+UnitreeG1:
+ algorithm: "VAIL"
+ algorithm_config:
+ std_0: 0.5
+ info_constraint: 0.1
+ lr_beta: 0.00001
+ z_dim: 128
+ disc_only_states: True
+ disc_use_next_states: False
+ train_disc_n_th_epoch: 3
+ disc_batch_size: 2048
+ learning_rate_critic: 1.0e-4
+ learning_rate_disc: 5.0e-5
+ policy_entr_coef: 1.0e-3
+ max_kl: 5.0e-3
+ n_epochs_cg: 25
+ use_noisy_targets: False
+ last_policy_activation: "identity"
+
 UnitreeA1.simple:
  algorithm: "VAIL"
  algorithm_config:

examples/imitation_learning/launcher.py

@@ -31,6 +31,7 @@
  env_ids = ["Atlas.walk", "Atlas.carry",
  "Talos.walk", "Talos.carry",
  "UnitreeH1.walk", "UnitreeH1.run", "UnitreeH1.carry",
+ "UnitreeG1.walk", "UnitreeG1.run",
  "HumanoidTorque.walk", "HumanoidTorque.run",
  "HumanoidMuscle.walk", "HumanoidMuscle.run",
  "UnitreeA1.simple", "UnitreeA1.hard"]

examples/replay_datasets/record_all.py

@@ -11,12 +11,13 @@ def experiment(seed=0):
  envs = LocoEnv.get_all_task_names()
 
  for env in envs:
+
  if "real" in env:
  replay_params = dict(n_episodes=15, n_steps_per_episode=250, record=True) if "Unitree.hard" in env or \
  ".all" in env or ".carry" in env \
  else dict(n_episodes=3, n_steps_per_episode=500, record=True)
 
- if ("Humanoid" in env or "Atlas.walk" in env) and not "carry" in env:
+ if ("Humanoid" in env or "Atlas.walk" in env or "UnitreeG1" in env) and not "carry" in env:
  env_params = dict(disable_arms=False)
  else:
  env_params = dict()

examples/replay_datasets/replay_Unitree_G1.py

@@ -0,0 +1,16 @@
+import numpy as np
+
+from loco_mujoco import LocoEnv
+
+
+def experiment(seed=0):
+
+ np.random.seed(seed)
+
+ mdp = LocoEnv.make("UnitreeG1.run")
+
+ mdp.play_trajectory_from_velocity(n_episodes=3, n_steps_per_episode=500)
+
+
+if __name__ == '__main__':
+ experiment()

examples/simple_mushroom_env/example_unitree_g1.py

@@ -0,0 +1,25 @@
+import numpy as np
+from loco_mujoco import LocoEnv
+
+# create the environment and task
+env = LocoEnv.make("UnitreeG1.walk")
+
+# get the dataset for the chosen environment and task
+expert_data = env.create_dataset()
+
+action_dim = env.info.action_space.shape[0]
+
+env.reset()
+env.render()
+absorbing = False
+i = 0
+
+while True:
+ if i == 1000 or absorbing:
+ env.reset()
+ i = 0
+ action = np.random.randn(action_dim)
+ nstate, reward, absorbing, info = env.step(action)
+
+ env.render()
+ i += 1

loco_mujoco/__init__.py

@@ -1,4 +1,4 @@
-__version__ = '0.2.2'
+__version__ = '0.3.0'
 
 try:
 

loco_mujoco/environments/README.md

@@ -9,12 +9,14 @@ The status of a dataset can be seen down below. ✅ means it is already availabl
  | ![out9](https://github.com/robfiras/loco-mujoco/assets/69359729/2c7aeb58-65a6-427c-8b12-197c96410cd8) | HumanoidMuscle4Ages.walk.2 </br> HumanoidMuscle4Ages.run.2 | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶 | Task of a (~5-6 year old) **Muscle** Humanoid Walking or Running Straight. |
  | ![out9](https://github.com/robfiras/loco-mujoco/assets/69359729/600a917f-c784-4b5e-ac99-9472711de843) | HumanoidMuscle4Ages.walk.1 </br> HumanoidMuscle4Ages.run.1 | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶 | Task of a (~2 year old) **Muscle** Humanoid Walking or Running Straight. |
  | ![out39](https://github.com/robfiras/loco-mujoco/assets/69359729/cf32520f-5a2e-401a-9f2e-b8033ef7109c) | HumanoidTorque.walk </br> HumanoidTorque.run </br> HumanoidTorque4Ages.walk.4 </br> HumanoidTorque.run </br> HumanoidTorque4Ages.run.4 | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶| Task of an adult **Torque** Humanoid Walking or Running Straight. |
- | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/352f3594-8903-4eaf-a223-f751b590f4ec) | HumanoidTorque4Ages.walk.3 </br> HumanoidTorque4Ages.run.3 | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶 | Task of a (~12 year old) **Torque** Humanoid Walking or Running Straight. |
- | ![out5](https://github.com/robfiras/loco-mujoco/assets/69359729/06c83af9-3c45-43a1-8173-aa1d8771fe4c) | HumanoidTorque4Ages.walk.2 </br> HumanoidTorque4Ages.run.2 | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶 | Task of a (~5-6 year old) **Torque** Humanoid Walking or Running Straight. |
- | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/5ec93baa-bed8-4d9f-b983-3bc12264b9b6) | HumanoidTorque4Ages.walk.1 </br> HumanoidTorque4Ages.run.1 | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶 | Task of a (~2 year old) **Torque** Humanoid Walking or Running Straight. |
+ | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/352f3594-8903-4eaf-a223-f751b590f4ec) | HumanoidTorque4Ages.walk.3 </br> HumanoidTorque4Ages.run.3 | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶 | Task of a (~12 year old) **Torque** Humanoid Walking or Running Straight. |
+ | ![out5](https://github.com/robfiras/loco-mujoco/assets/69359729/06c83af9-3c45-43a1-8173-aa1d8771fe4c) | HumanoidTorque4Ages.walk.2 </br> HumanoidTorque4Ages.run.2 | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶 | Task of a (~5-6 year old) **Torque** Humanoid Walking or Running Straight. |
+ | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/5ec93baa-bed8-4d9f-b983-3bc12264b9b6) | HumanoidTorque4Ages.walk.1 </br> HumanoidTorque4Ages.run.1 | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶 | Task of a (~2 year old) **Torque** Humanoid Walking or Running Straight. |
  | ![out8](https://github.com/robfiras/loco-mujoco/assets/69359729/fed0315c-921e-4b2e-a9c2-54b85198ef65) | UnitreeH1.walk | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶| UnitreeH1 Straight Walking Task. |
  | ![out9](https://github.com/robfiras/loco-mujoco/assets/69359729/ab0dec59-fc24-4763-8ff6-38d58ac3b3de) | UnitreeH1.run | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶| UnitreeH1 Straight Running Task. |
  | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/851ff3c0-d05f-4de1-a00b-7b3204056e2f) | UnitreeH1.carry | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶| UnitreeH1 Carry Task. |
+ | ![out8](https://github.com/robfiras/loco-mujoco/assets/69359729/e9a33e94-65e4-48f5-8d7a-3d103c4ec7f5) | UnitreeG1.walk | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶| UnitreeG1 Straight Walking Task. |
+ | ![out9](https://github.com/robfiras/loco-mujoco/assets/69359729/57292ef1-c152-4cd3-b7c1-d9a1e0633481) | UnitreeG1.run | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶| UnitreeG1 Straight Running Task. |
  | ![out7](https://github.com/robfiras/loco-mujoco/assets/69359729/22c7bb0c-ff92-4e99-a964-7654df6d22c4) | Talos.walk | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶| Talos Straight Walking Task. |
  | ![out3](https://github.com/robfiras/loco-mujoco/assets/69359729/0ba1f0e7-1f3d-4088-a44f-0a53bec1cf3a) | Talos.carry | real: ✅ <br /> perfect: 🔶 <br /> preference: 🔶| Talos Carry Task. |
  | ![out13](https://github.com/robfiras/loco-mujoco/assets/69359729/1ff09d98-e46b-429c-ac07-87de58853d28) | Atlas.walk | real: ✅ <br /> perfect: ✅ <br /> preference: 🔶| Atlas Straight Walking Task. |

loco_mujoco/environments/base.py

@@ -37,7 +37,7 @@ def __init__(self, xml_handles, action_spec, observation_spec, collision_groups=
  Constructor.
 
  Args:
- xml_handle : MuJoCo xml handle.
+ xml_handles : MuJoCo xml handles.
  actuation_spec (list): A list specifying the names of the joints
  which should be controllable by the agent. Can be left empty
  when all actuators should be used;
@@ -86,6 +86,7 @@ def __init__(self, xml_handles, action_spec, observation_spec, collision_groups=
 
  if type(xml_handles) != list:
  xml_handles = [xml_handles]
+ self._xml_handles = xml_handles
 
  if collision_groups is None:
  collision_groups = list()
@@ -128,7 +129,7 @@ def __init__(self, xml_handles, action_spec, observation_spec, collision_groups=
  self.mean_grf = self._setup_ground_force_statistics()
 
  # dataset dummy
- self._dataset= None
+ self._dataset = None
 
  if traj_params:
  self.trajectories = None
@@ -547,6 +548,22 @@ def load_dataset_and_get_traj_files(self, dataset_path, freq=None):
 
  return trajectories
 
+ @property
+ def xml_handle(self):
+ """ Returns the XML handle of the environment. This will raise an error if the environment contains more
+ than one xml_handle. """
+
+ if len(self._xml_handles) > 1:
+ raise ValueError("This environment contains multiple models and hence multiple xml_handles. Use the "
+ "property \"xml_handles\" instead.")
+ return self._xml_handles[0]
+
+ @property
+ def xml_handles(self):
+ """ Returns all XML handles of the environment. """
+
+ return self._xml_handles
+
  def _get_observation_space(self):
  """
  Returns a tuple of the lows and highs (np.array) of the observation space.