Merge remote-tracking branch 'origin/dmc_integration' into dmc_integration

# Conflicts:
#	README.md
#	alr_envs/__init__.py
#	setup.py

README.md

@@ -1,12 +1,11 @@
-## ALR Environments
+## ALR Robotics Control Environments
     
 This repository collects custom Robotics environments not included in benchmark suites like OpenAI gym, rllab, etc. 
 Creating a custom (Mujoco) gym environment can be done according to [this guide](https://github.com/openai/gym/blob/master/docs/creating-environments.md).
 For stochastic search problems with gym interface use the `Rosenbrock-v0` reference implementation.
-We also support to solve environments with DMPs. When adding new DMP tasks check the `ViaPointReacherDMP-v0` reference implementation.
-When simply using the tasks, you can also leverage the wrapper class `DmpWrapper` to turn normal gym environments in to DMP tasks.
+We also support to solve environments with Dynamic Movement Primitives (DMPs) and Probabilistic Movement Primitives (DetPMP, we only consider the mean usually). 
 
-## Environments
+## Step-based Environments
 Currently we have the following environments: 
 
 ### Mujoco
@@ -32,11 +31,13 @@ Currently we have the following environments:
 |`ViaPointReacher-v0`| Simple reaching task leveraging a via point, which supports self collision detection. Provides a reward only at 100 and 199 for reaching the viapoint and goal point, respectively.| 200 | 5 | 18 
 |`HoleReacher-v0`| 5 link reaching task where the end-effector needs to reach into a narrow hole without collding with itself or walls | 200 | 5 | 18
 
-### DMP Environments
-These environments are closer to stochastic search. They always execute a full trajectory, which is computed by a DMP and executed by a controller, e.g. a PD controller.
-The goal is to learn the parameters of this DMP to generate a suitable trajectory. 
-All environments provide the full episode reward and additional information about early terminations, e.g. due to collisions. 
+## Motion Primitive Environments (Episodic environments)
+Unlike step-based environments, these motion primitive (MP) environments are closer to stochastic search and what can be found in robotics. They always execute a full trajectory, which is computed by a Dynamic Motion Primitive (DMP) or Probabilitic Motion Primitive (DetPMP) and translated into individual actions with a controller, e.g. a PD controller. The actual Controller, however, depends on the type of environment, i.e. position, velocity, or torque controlled.
+The goal is to learn the parametrization of the motion primitives in order to generate a suitable trajectory. 
+MP This can also be done in a contextual setting, where all changing elements of the task are exposed once in the beginning. This requires to find a new parametrization for each trajectory.
+All environments provide the full cumulative episode reward and additional information about early terminations, e.g. due to collisions. 
 
+### Classic Control
 |Name| Description|Horizon|Action Dimension|Context Dimension
 |---|---|---|---|---|
 |`ViaPointReacherDMP-v0`| A DMP provides a trajectory for the `ViaPointReacher-v0` task. | 200 | 25
@@ -48,6 +49,29 @@ All environments provide the full episode reward and additional information abou
 
 [//]:  |`HoleReacherDetPMP-v0`|
 
+### OpenAI gym Environments
+These environments are wrapped-versions of their OpenAI-gym counterparts.
+
+|Name| Description|Trajectory Horizon|Action Dimension|Context Dimension
+|---|---|---|---|---|
+|`ContinuousMountainCarDetPMP-v0`| A DetPmP wrapped version of the ContinuousMountainCar-v0 environment. | 100 | 1
+|`ReacherDetPMP-v2`| A DetPmP wrapped version of the Reacher-v2 environment. | 50 | 2
+|`FetchSlideDenseDetPMP-v1`| A DetPmP wrapped version of the FetchSlideDense-v1 environment. | 50 | 4 
+|`FetchReachDenseDetPMP-v1`| A DetPmP wrapped version of the FetchReachDense-v1 environment. | 50 | 4
+
+### Deep Mind Control Suite Environments
+These environments are wrapped-versions of their Deep Mind Control Suite (DMC) counterparts.
+Given most task can be solved in shorter horizon lengths than the original 1000 steps, we often shorten the episodes for those task. 
+
+|Name| Description|Trajectory Horizon|Action Dimension|Context Dimension
+|---|---|---|---|---|
+|`dmc_ball_in_cup-catch_detpmp-v0`| A DetPmP wrapped version of the "catch" task for the "ball_in_cup" environment. | 50 | 10 | 2
+|`dmc_ball_in_cup-catch_dmp-v0`| A DMP wrapped version of the "catch" task for the "ball_in_cup" environment. | 50| 10 | 2
+|`dmc_reacher-easy_detpmp-v0`| A DetPmP wrapped version of the "easy" task for the "reacher" environment. | 1000 | 10 | 4
+|`dmc_reacher-easy_dmp-v0`| A DMP wrapped version of the "easy" task for the "reacher" environment. | 1000| 10 | 4
+|`dmc_reacher-hard_detpmp-v0`| A DetPmP wrapped version of the "hard" task for the "reacher" environment.| 1000 | 10 | 4
+|`dmc_reacher-hard_dmp-v0`| A DMP wrapped version of the "hard" task for the "reacher" environment. | 1000 | 10 | 4
+
 ## Install
 1. Clone the repository 
 ```bash 
@@ -78,4 +102,4 @@ for i in range(10000):
 
 ``` 
 
-For an example using a DMP wrapped env and asynchronous sampling look at [mp_env_async_sampler.py](./alr_envs/utils/mp_env_async_sampler.py)
+For an example using a DMP wrapped env and asynchronous sampling look at [mp_env_async_sampler.py](./alr_envs/utils/mp_env_async_sampler.py)

alr_envs/__init__.py

@@ -8,6 +8,7 @@ from alr_envs.dmc.manipulation.reach.reach_mp_wrapper import DMCReachSiteMPWrapp
 from alr_envs.dmc.suite.ball_in_cup.ball_in_cup_mp_wrapper import DMCBallInCupMPWrapper
 from alr_envs.dmc.suite.cartpole.cartpole_mp_wrapper import DMCCartpoleMPWrapper, DMCCartpoleThreePolesMPWrapper, \
     DMCCartpoleTwoPolesMPWrapper
+from alr_envs.open_ai import reacher_v2, continuous_mountain_car, fetch
 from alr_envs.dmc.suite.reacher.reacher_mp_wrapper import DMCReacherMPWrapper
 
 # Mujoco
@@ -790,3 +791,80 @@ register(
         }
     }
 )
+
+## Open AI
+register(
+    id='ContinuousMountainCarDetPMP-v0',
+    entry_point='alr_envs.utils.make_env_helpers:make_detpmp_env_helper',
+    kwargs={
+        "name": "gym.envs.classic_control:MountainCarContinuous-v0",
+        "wrappers": [continuous_mountain_car.MPWrapper],
+        "mp_kwargs": {
+            "num_dof": 1,
+            "num_basis": 4,
+            "duration": 2,
+            "post_traj_time": 0,
+            "width": 0.02,
+            "policy_type": "motor",
+            "policy_kwargs": {
+                "p_gains": 1.,
+                "d_gains": 1.
+            }
+        }
+    }
+)
+
+register(
+    id='ReacherDetPMP-v2',
+    entry_point='alr_envs.utils.make_env_helpers:make_detpmp_env_helper',
+    kwargs={
+        "name": "gym.envs.mujoco:Reacher-v2",
+        "wrappers": [reacher_v2.MPWrapper],
+        "mp_kwargs": {
+            "num_dof": 2,
+            "num_basis": 6,
+            "duration": 1,
+            "post_traj_time": 0,
+            "width": 0.02,
+            "policy_type": "motor",
+            "policy_kwargs": {
+                "p_gains": .6,
+                "d_gains": .075
+            }
+        }
+    }
+)
+
+register(
+    id='FetchSlideDenseDetPMP-v1',
+    entry_point='alr_envs.utils.make_env_helpers:make_detpmp_env_helper',
+    kwargs={
+        "name": "gym.envs.robotics:FetchSlideDense-v1",
+        "wrappers": [fetch.MPWrapper],
+        "mp_kwargs": {
+            "num_dof": 4,
+            "num_basis": 5,
+            "duration": 2,
+            "post_traj_time": 0,
+            "width": 0.02,
+            "policy_type": "position"
+        }
+    }
+)
+
+register(
+    id='FetchReachDenseDetPMP-v1',
+    entry_point='alr_envs.utils.make_env_helpers:make_detpmp_env_helper',
+    kwargs={
+        "name": "gym.envs.robotics:FetchReachDense-v1",
+        "wrappers": [fetch.MPWrapper],
+        "mp_kwargs": {
+            "num_dof": 4,
+            "num_basis": 5,
+            "duration": 2,
+            "post_traj_time": 0,
+            "width": 0.02,
+            "policy_type": "position"
+        }
+    }
+)

alr_envs/examples/examples_open_ai.py

@@ -0,0 +1,41 @@
+from alr_envs.utils.make_env_helpers import make_env
+
+
+def example_mp(env_name, seed=1):
+    """
+    Example for running a motion primitive based version of a OpenAI-gym environment, which is already registered.
+    For more information on motion primitive specific stuff, look at the mp examples.
+    Args:
+        env_name: DetPMP env_id
+        seed: seed
+
+    Returns:
+
+    """
+    # While in this case gym.make() is possible to use as well, we recommend our custom make env function.
+    env = make_env(env_name, seed)
+
+    rewards = 0
+    obs = env.reset()
+
+    # number of samples/full trajectories (multiple environment steps)
+    for i in range(10):
+        ac = env.action_space.sample()
+        obs, reward, done, info = env.step(ac)
+        rewards += reward
+
+        if done:
+            print(rewards)
+            rewards = 0
+            obs = env.reset()
+
+if __name__ == '__main__':
+    # DMP - not supported yet
+    #example_mp("ReacherDetPMP-v2")
+
+    # DetProMP
+    example_mp("ContinuousMountainCarDetPMP-v0")
+    example_mp("ReacherDetPMP-v2")
+    example_mp("FetchReachDenseDetPMP-v1")
+    example_mp("FetchSlideDenseDetPMP-v1")
+

alr_envs/open_ai/continuous_mountain_car/__init__.py

@@ -0,0 +1 @@
+from alr_envs.open_ai.continuous_mountain_car.mp_wrapper import MPWrapper

alr_envs/open_ai/continuous_mountain_car/mp_wrapper.py

@@ -0,0 +1,22 @@
+from typing import Union
+
+import numpy as np
+from mp_env_api.interface_wrappers.mp_env_wrapper import MPEnvWrapper
+
+
+class MPWrapper(MPEnvWrapper):
+    @property
+    def current_vel(self) -> Union[float, int, np.ndarray]:
+        return np.array([self.state[1]])
+
+    @property
+    def current_pos(self) -> Union[float, int, np.ndarray]:
+        return np.array([self.state[0]])
+
+    @property
+    def goal_pos(self):
+        raise ValueError("Goal position is not available and has to be learnt based on the environment.")
+
+    @property
+    def dt(self) -> Union[float, int]:
+        return 0.02

alr_envs/open_ai/fetch/__init__.py

@@ -0,0 +1 @@
+from alr_envs.open_ai.fetch.mp_wrapper import MPWrapper

alr_envs/open_ai/fetch/mp_wrapper.py

@@ -0,0 +1,22 @@
+from typing import Union
+
+import numpy as np
+from mp_env_api.interface_wrappers.mp_env_wrapper import MPEnvWrapper
+
+
+class MPWrapper(MPEnvWrapper):
+    @property
+    def current_vel(self) -> Union[float, int, np.ndarray]:
+        return self.unwrapped._get_obs()["observation"][-5:-1]
+
+    @property
+    def current_pos(self) -> Union[float, int, np.ndarray]:
+        return self.unwrapped._get_obs()["observation"][:4]
+
+    @property
+    def goal_pos(self):
+        raise ValueError("Goal position is not available and has to be learnt based on the environment.")
+
+    @property
+    def dt(self) -> Union[float, int]:
+        return self.env.dt

alr_envs/open_ai/reacher_v2/__init__.py

@@ -0,0 +1 @@
+from alr_envs.open_ai.reacher_v2.mp_wrapper import MPWrapper

alr_envs/open_ai/reacher_v2/mp_wrapper.py

@@ -0,0 +1,19 @@
+from typing import Union
+
+import numpy as np
+from mp_env_api.interface_wrappers.mp_env_wrapper import MPEnvWrapper
+
+
+class MPWrapper(MPEnvWrapper):
+
+    @property
+    def current_vel(self) -> Union[float, int, np.ndarray]:
+        return self.sim.data.qvel[:2]
+
+    @property
+    def current_pos(self) -> Union[float, int, np.ndarray]:
+        return self.sim.data.qpos[:2]
+
+    @property
+    def dt(self) -> Union[float, int]:
+        return self.env.dt

reacher.egg-info/PKG-INFO

@@ -1,10 +0,0 @@
-Metadata-Version: 1.0
-Name: reacher
-Version: 0.0.1
-Summary: UNKNOWN
-Home-page: UNKNOWN
-Author: UNKNOWN
-Author-email: UNKNOWN
-License: UNKNOWN
-Description: UNKNOWN
-Platform: UNKNOWN

reacher.egg-info/SOURCES.txt

@@ -1,7 +0,0 @@
-README.md
-setup.py
-reacher.egg-info/PKG-INFO
-reacher.egg-info/SOURCES.txt
-reacher.egg-info/dependency_links.txt
-reacher.egg-info/requires.txt
-reacher.egg-info/top_level.txt

reacher.egg-info/dependency_links.txt

@@ -1 +0,0 @@
-

reacher.egg-info/requires.txt

@@ -1 +0,0 @@
-gym

reacher.egg-info/top_level.txt

@@ -1 +0,0 @@
-

setup.py

@@ -3,14 +3,15 @@ from setuptools import setup
 setup(
     name='alr_envs',
     version='0.0.1',
-    packages=['alr_envs', 'alr_envs.classic_control', 'alr_envs.mujoco', 'alr_envs.stochastic_search',
+    packages=['alr_envs', 'alr_envs.classic_control', 'alr_envs.open_ai', 'alr_envs.mujoco', 'alr_envs.stochastic_search',
               'alr_envs.utils'],
     install_requires=[
         'gym',
         'PyQt5',
         'matplotlib',
-        # 'mp_env_api @ git+ssh://git@github.com/ALRhub/motion_primitive_env_api.git',
-        'mujoco_py'
+        'mp_env_api @ git+ssh://git@github.com/ALRhub/motion_primitive_env_api.git',
+        'mujoco-py<2.1,>=2.0',
+        'dm_control'
     ],
 
     url='https://github.com/ALRhub/alr_envs/',