renaming to fancy_gym
This commit is contained in:
Fabian
@@ -0,0 +1,137 @@
|
||||
import fancy_gym
|
||||
|
||||
|
||||
def example_dmc(env_id="dmc:fish-swim", seed=1, iterations=1000, render=True):
|
||||
"""
|
||||
Example for running a DMC based env in the step based setting.
|
||||
The env_id has to be specified as `domain_name:task_name` or
|
||||
for manipulation tasks as `domain_name:manipulation-environment_name`
|
||||
|
||||
Args:
|
||||
env_id: Either `domain_name-task_name` or `manipulation-environment_name`
|
||||
seed: seed for deterministic behaviour
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
env = fancy_gym.make(env_id, seed)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
print("observation shape:", env.observation_space.shape)
|
||||
print("action shape:", env.action_space.shape)
|
||||
|
||||
for i in range(iterations):
|
||||
ac = env.action_space.sample()
|
||||
if render:
|
||||
env.render(mode="human")
|
||||
obs, reward, done, info = env.step(ac)
|
||||
rewards += reward
|
||||
|
||||
if done:
|
||||
print(env_id, rewards)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
env.close()
|
||||
del env
|
||||
|
||||
|
||||
def example_custom_dmc_and_mp(seed=1, iterations=1, render=True):
|
||||
"""
|
||||
Example for running a custom movement primitive based environments.
|
||||
Our already registered environments follow the same structure.
|
||||
Hence, this also allows to adjust hyperparameters of the movement primitives.
|
||||
Yet, we recommend the method above if you are just interested in chaining those parameters for existing tasks.
|
||||
We appreciate PRs for custom environments (especially MP wrappers of existing tasks)
|
||||
for our repo: https://github.com/ALRhub/fancy_gym/
|
||||
Args:
|
||||
seed: seed for deterministic behaviour
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
|
||||
# Base DMC name, according to structure of above example
|
||||
base_env_id = "dmc:ball_in_cup-catch"
|
||||
|
||||
# Replace this wrapper with the custom wrapper for your environment by inheriting from the RawInterfaceWrapper.
|
||||
# You can also add other gym.Wrappers in case they are needed.
|
||||
wrappers = [fancy_gym.dmc.suite.ball_in_cup.MPWrapper]
|
||||
# # For a ProMP
|
||||
trajectory_generator_kwargs = {'trajectory_generator_type': 'promp'}
|
||||
phase_generator_kwargs = {'phase_generator_type': 'linear'}
|
||||
controller_kwargs = {'controller_type': 'motor',
|
||||
"p_gains": 1.0,
|
||||
"d_gains": 0.1,}
|
||||
basis_generator_kwargs = {'basis_generator_type': 'zero_rbf',
|
||||
'num_basis': 5,
|
||||
'num_basis_zero_start': 1
|
||||
}
|
||||
|
||||
# For a DMP
|
||||
# trajectory_generator_kwargs = {'trajectory_generator_type': 'dmp'}
|
||||
# phase_generator_kwargs = {'phase_generator_type': 'exp',
|
||||
# 'alpha_phase': 2}
|
||||
# controller_kwargs = {'controller_type': 'motor',
|
||||
# "p_gains": 1.0,
|
||||
# "d_gains": 0.1,
|
||||
# }
|
||||
# basis_generator_kwargs = {'basis_generator_type': 'rbf',
|
||||
# 'num_basis': 5
|
||||
# }
|
||||
env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
|
||||
traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
|
||||
phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
|
||||
seed=seed)
|
||||
|
||||
# This renders the full MP 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.
|
||||
if render:
|
||||
env.render(mode="human")
|
||||
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(iterations):
|
||||
ac = env.action_space.sample()
|
||||
obs, reward, done, info = env.step(ac)
|
||||
rewards += reward
|
||||
|
||||
if done:
|
||||
print(base_env_id, rewards)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
env.close()
|
||||
del env
|
||||
|
||||
|
||||
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.
|
||||
|
||||
# For rendering DMC
|
||||
# export MUJOCO_GL="osmesa"
|
||||
render = True
|
||||
|
||||
# # Standard DMC Suite tasks
|
||||
example_dmc("dmc:fish-swim", seed=10, iterations=1000, render=render)
|
||||
#
|
||||
# # Manipulation tasks
|
||||
# # Disclaimer: The vision versions are currently not integrated and yield an error
|
||||
example_dmc("dmc:manipulation-reach_site_features", seed=10, iterations=250, render=render)
|
||||
#
|
||||
# # Gym + DMC hybrid task provided in the MP framework
|
||||
example_dmc("dmc_ball_in_cup-catch_promp-v0", seed=10, iterations=1, render=render)
|
||||
|
||||
# Custom DMC task # Different seed, because the episode is longer for this example and the name+seed combo is
|
||||
# already registered above
|
||||
example_custom_dmc_and_mp(seed=11, iterations=1, render=render)
|
||||
@@ -0,0 +1,100 @@
|
||||
from collections import defaultdict
|
||||
|
||||
import gym
|
||||
import numpy as np
|
||||
|
||||
import fancy_gym
|
||||
|
||||
|
||||
def example_general(env_id="Pendulum-v1", seed=1, iterations=1000, render=True):
|
||||
"""
|
||||
Example for running any env in the step based setting.
|
||||
This also includes DMC environments when leveraging our custom make_env function.
|
||||
|
||||
Args:
|
||||
env_id: OpenAI/Custom gym task id or either `domain_name-task_name` or `manipulation-environment_name` for DMC tasks
|
||||
seed: seed for deterministic behaviour
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
|
||||
env = fancy_gym.make(env_id, seed)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
print("Observation shape: ", env.observation_space.shape)
|
||||
print("Action shape: ", env.action_space.shape)
|
||||
|
||||
# number of environment steps
|
||||
for i in range(iterations):
|
||||
obs, reward, done, info = env.step(env.action_space.sample())
|
||||
rewards += reward
|
||||
|
||||
if render:
|
||||
env.render()
|
||||
|
||||
if done:
|
||||
print(rewards)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
|
||||
def example_async(env_id="HoleReacher-v0", n_cpu=4, seed=int('533D', 16), n_samples=800):
|
||||
"""
|
||||
Example for running any env in a vectorized multiprocessing setting to generate more samples faster.
|
||||
This also includes DMC and DMP environments when leveraging our custom make_env function.
|
||||
Be aware, increasing the number of environments reduces the total length of the individual episodes.
|
||||
|
||||
Args:
|
||||
env_id: OpenAI/Custom gym task id or either `domain_name-task_name` or `manipulation-environment_name` for DMC tasks
|
||||
seed: seed for deterministic behaviour
|
||||
n_cpu: Number of cpus cores to use in parallel
|
||||
n_samples: number of samples generated in total by all environments.
|
||||
|
||||
Returns: Tuple of (obs, reward, done, info) with type np.ndarray
|
||||
|
||||
"""
|
||||
env = gym.vector.AsyncVectorEnv([fancy_gym.make_rank(env_id, seed, i) for i in range(n_cpu)])
|
||||
# OR
|
||||
# envs = gym.vector.AsyncVectorEnv([make_env(env_id, seed + i) for i in range(n_cpu)])
|
||||
|
||||
# for plotting
|
||||
rewards = np.zeros(n_cpu)
|
||||
buffer = defaultdict(list)
|
||||
|
||||
obs = env.reset()
|
||||
|
||||
# this would generate more samples than requested if n_samples % num_envs != 0
|
||||
repeat = int(np.ceil(n_samples / env.num_envs))
|
||||
for i in range(repeat):
|
||||
obs, reward, done, info = env.step(env.action_space.sample())
|
||||
buffer['obs'].append(obs)
|
||||
buffer['reward'].append(reward)
|
||||
buffer['done'].append(done)
|
||||
buffer['info'].append(info)
|
||||
rewards += reward
|
||||
if np.any(done):
|
||||
print(f"Reward at iteration {i}: {rewards[done]}")
|
||||
rewards[done] = 0
|
||||
|
||||
# do not return values above threshold
|
||||
return *map(lambda v: np.stack(v)[:n_samples], buffer.values()),
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
render = True
|
||||
|
||||
# Basic gym task
|
||||
example_general("Pendulum-v1", seed=10, iterations=200, render=render)
|
||||
|
||||
# Mujoco task from framework
|
||||
example_general("Reacher5d-v0", seed=10, iterations=200, render=render)
|
||||
|
||||
# # OpenAI Mujoco task
|
||||
example_general("HalfCheetah-v2", seed=10, render=render)
|
||||
|
||||
# Vectorized multiprocessing environments
|
||||
# example_async(env_id="HoleReacher-v0", n_cpu=2, seed=int('533D', 16), n_samples=2 * 200)
|
||||
|
||||
@@ -0,0 +1,134 @@
|
||||
import fancy_gym
|
||||
|
||||
|
||||
def example_dmc(env_id="fish-swim", seed=1, iterations=1000, render=True):
|
||||
"""
|
||||
Example for running a MetaWorld based env in the step based setting.
|
||||
The env_id has to be specified as `task_name-v2`. V1 versions are not supported and we always
|
||||
return the observable goal version.
|
||||
All tasks can be found here: https://arxiv.org/pdf/1910.10897.pdf or https://meta-world.github.io/
|
||||
|
||||
Args:
|
||||
env_id: `task_name-v2`
|
||||
seed: seed for deterministic behaviour (TODO: currently not working due to an issue in MetaWorld code)
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
env = fancy_gym.make(env_id, seed)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
print("observation shape:", env.observation_space.shape)
|
||||
print("action shape:", env.action_space.shape)
|
||||
|
||||
for i in range(iterations):
|
||||
ac = env.action_space.sample()
|
||||
if render:
|
||||
# THIS NEEDS TO BE SET TO FALSE FOR NOW, BECAUSE THE INTERFACE FOR RENDERING IS DIFFERENT TO BASIC GYM
|
||||
# TODO: Remove this, when Metaworld fixes its interface.
|
||||
env.render(False)
|
||||
obs, reward, done, info = env.step(ac)
|
||||
rewards += reward
|
||||
if done:
|
||||
print(env_id, rewards)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
env.close()
|
||||
del env
|
||||
|
||||
|
||||
def example_custom_dmc_and_mp(seed=1, iterations=1, render=True):
|
||||
"""
|
||||
Example for running a custom movement primitive based environments.
|
||||
Our already registered environments follow the same structure.
|
||||
Hence, this also allows to adjust hyperparameters of the movement primitives.
|
||||
Yet, we recommend the method above if you are just interested in chaining those parameters for existing tasks.
|
||||
We appreciate PRs for custom environments (especially MP wrappers of existing tasks)
|
||||
for our repo: https://github.com/ALRhub/fancy_gym/
|
||||
Args:
|
||||
seed: seed for deterministic behaviour (TODO: currently not working due to an issue in MetaWorld code)
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode (TODO: currently not working due to an issue in MetaWorld code)
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
|
||||
# Base MetaWorld name, according to structure of above example
|
||||
base_env_id = "metaworld:button-press-v2"
|
||||
|
||||
# Replace this wrapper with the custom wrapper for your environment by inheriting from the RawInterfaceWrapper.
|
||||
# You can also add other gym.Wrappers in case they are needed.
|
||||
wrappers = [fancy_gym.meta.goal_object_change_mp_wrapper.MPWrapper]
|
||||
# # For a ProMP
|
||||
# trajectory_generator_kwargs = {'trajectory_generator_type': 'promp'}
|
||||
# phase_generator_kwargs = {'phase_generator_type': 'linear'}
|
||||
# controller_kwargs = {'controller_type': 'metaworld'}
|
||||
# basis_generator_kwargs = {'basis_generator_type': 'zero_rbf',
|
||||
# 'num_basis': 5,
|
||||
# 'num_basis_zero_start': 1
|
||||
# }
|
||||
|
||||
# For a DMP
|
||||
trajectory_generator_kwargs = {'trajectory_generator_type': 'dmp'}
|
||||
phase_generator_kwargs = {'phase_generator_type': 'exp',
|
||||
'alpha_phase': 2}
|
||||
controller_kwargs = {'controller_type': 'metaworld'}
|
||||
basis_generator_kwargs = {'basis_generator_type': 'rbf',
|
||||
'num_basis': 5
|
||||
}
|
||||
env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
|
||||
traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
|
||||
phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
|
||||
seed=seed)
|
||||
|
||||
# This renders the full MP 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.
|
||||
if render:
|
||||
raise ValueError("Metaworld render interface bug does not allow to render() fixes its interface. "
|
||||
"A temporary workaround is to alter their code in MujocoEnv render() from "
|
||||
"`if not offscreen` to `if not offscreen or offscreen == 'human'`.")
|
||||
# TODO: Remove this, when Metaworld fixes its interface.
|
||||
# env.render(mode="human")
|
||||
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(iterations):
|
||||
ac = env.action_space.sample()
|
||||
obs, reward, done, info = env.step(ac)
|
||||
rewards += reward
|
||||
|
||||
if done:
|
||||
print(base_env_id, rewards)
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
env.close()
|
||||
del env
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
# Disclaimer: MetaWorld environments require the seed to be specified in the beginning.
|
||||
# Adjusting it afterwards with env.seed() is not recommended as it may not affect the underlying behavior.
|
||||
|
||||
# For rendering it might be necessary to specify your OpenGL installation
|
||||
# export LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libGLEW.so
|
||||
render = False
|
||||
|
||||
# # Standard Meta world tasks
|
||||
example_dmc("metaworld:button-press-v2", seed=10, iterations=500, render=render)
|
||||
|
||||
# # MP + MetaWorld hybrid task provided in the our framework
|
||||
example_dmc("ButtonPressProMP-v2", seed=10, iterations=1, render=render)
|
||||
#
|
||||
# # Custom MetaWorld task
|
||||
example_custom_dmc_and_mp(seed=10, iterations=1, render=render)
|
||||
|
||||
@@ -0,0 +1,169 @@
|
||||
import fancy_gym
|
||||
|
||||
|
||||
def example_mp(env_name="HoleReacherProMP-v0", seed=1, iterations=1, render=True):
|
||||
"""
|
||||
Example for running a black box based environment, which is already registered
|
||||
Args:
|
||||
env_name: Black box env_id
|
||||
seed: seed for deterministic behaviour
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
# Equivalent to gym, we have a make function which can be used to create environments.
|
||||
# It takes care of seeding and enables the use of a variety of external environments using the gym interface.
|
||||
env = fancy_gym.make(env_name, seed)
|
||||
|
||||
returns = 0
|
||||
# env.render(mode=None)
|
||||
obs = env.reset()
|
||||
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(iterations):
|
||||
|
||||
if render and i % 2 == 0:
|
||||
# This renders the full MP 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 the mode multiple times when
|
||||
# e.g. only every second trajectory should be displayed, such as here
|
||||
# Just make sure the correct mode is set before executing the step.
|
||||
env.render(mode="human")
|
||||
else:
|
||||
env.render(mode=None)
|
||||
|
||||
# Now the action space is not the raw action but the parametrization of the trajectory generator,
|
||||
# such as a ProMP
|
||||
ac = env.action_space.sample()
|
||||
# This executes a full trajectory and gives back the context (obs) of the last step in the trajectory, or the
|
||||
# full observation space of the last step, if replanning/sub-trajectory learning is used. The 'reward' is equal
|
||||
# to the return of a trajectory. Default is the sum over the step-wise rewards.
|
||||
obs, reward, done, info = env.step(ac)
|
||||
# Aggregated returns
|
||||
returns += reward
|
||||
|
||||
if done:
|
||||
print(reward)
|
||||
obs = env.reset()
|
||||
|
||||
|
||||
def example_custom_mp(env_name="Reacher5dProMP-v0", seed=1, iterations=1, render=True):
|
||||
"""
|
||||
Example for running a movement primitive based environment, which is already registered
|
||||
Args:
|
||||
env_name: DMP env_id
|
||||
seed: seed for deterministic behaviour
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
# Changing the arguments of the black box env is possible by providing them to gym as with all kwargs.
|
||||
# E.g. here for way to many basis functions
|
||||
env = fancy_gym.make(env_name, seed, basis_generator_kwargs={'num_basis': 1000})
|
||||
# env = fancy_gym.make(env_name, seed)
|
||||
# mp_dict.update({'black_box_kwargs': {'learn_sub_trajectories': True}})
|
||||
# mp_dict.update({'black_box_kwargs': {'do_replanning': lambda pos, vel, t: lambda t: t % 100}})
|
||||
|
||||
returns = 0
|
||||
obs = env.reset()
|
||||
|
||||
# This time rendering every trajectory
|
||||
if render:
|
||||
env.render(mode="human")
|
||||
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(iterations):
|
||||
ac = env.action_space.sample()
|
||||
obs, reward, done, info = env.step(ac)
|
||||
returns += reward
|
||||
|
||||
if done:
|
||||
print(i, reward)
|
||||
obs = env.reset()
|
||||
|
||||
return obs
|
||||
|
||||
|
||||
def example_fully_custom_mp(seed=1, iterations=1, render=True):
|
||||
"""
|
||||
Example for running a custom movement primitive based environments.
|
||||
Our already registered environments follow the same structure.
|
||||
Hence, this also allows to adjust hyperparameters of the movement primitives.
|
||||
Yet, we recommend the method above if you are just interested in changing those parameters for existing tasks.
|
||||
We appreciate PRs for custom environments (especially MP wrappers of existing tasks)
|
||||
for our repo: https://github.com/ALRhub/fancy_gym/
|
||||
Args:
|
||||
seed: seed
|
||||
iterations: Number of rollout steps to run
|
||||
render: Render the episode
|
||||
|
||||
Returns:
|
||||
|
||||
"""
|
||||
|
||||
base_env_id = "HoleReacher-v0"
|
||||
|
||||
# Replace this wrapper with the custom wrapper for your environment by inheriting from the RawInterfaceWrapper.
|
||||
# You can also add other gym.Wrappers in case they are needed.
|
||||
wrappers = [fancy_gym.envs.classic_control.hole_reacher.MPWrapper]
|
||||
|
||||
# # For a ProMP
|
||||
# trajectory_generator_kwargs = {'trajectory_generator_type': 'promp',
|
||||
# 'weight_scale': 2}
|
||||
# phase_generator_kwargs = {'phase_generator_type': 'linear'}
|
||||
# controller_kwargs = {'controller_type': 'velocity'}
|
||||
# basis_generator_kwargs = {'basis_generator_type': 'zero_rbf',
|
||||
# 'num_basis': 5,
|
||||
# 'num_basis_zero_start': 1
|
||||
# }
|
||||
|
||||
# For a DMP
|
||||
trajectory_generator_kwargs = {'trajectory_generator_type': 'dmp',
|
||||
'weight_scale': 500}
|
||||
phase_generator_kwargs = {'phase_generator_type': 'exp',
|
||||
'alpha_phase': 2.5}
|
||||
controller_kwargs = {'controller_type': 'velocity'}
|
||||
basis_generator_kwargs = {'basis_generator_type': 'rbf',
|
||||
'num_basis': 5
|
||||
}
|
||||
env = fancy_gym.make_bb(env_id=base_env_id, wrappers=wrappers, black_box_kwargs={},
|
||||
traj_gen_kwargs=trajectory_generator_kwargs, controller_kwargs=controller_kwargs,
|
||||
phase_kwargs=phase_generator_kwargs, basis_kwargs=basis_generator_kwargs,
|
||||
seed=seed)
|
||||
|
||||
if render:
|
||||
env.render(mode="human")
|
||||
|
||||
rewards = 0
|
||||
obs = env.reset()
|
||||
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(iterations):
|
||||
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__':
|
||||
render = True
|
||||
# DMP
|
||||
example_mp("HoleReacherDMP-v0", seed=10, iterations=5, render=render)
|
||||
#
|
||||
# # ProMP
|
||||
example_mp("HoleReacherProMP-v0", seed=10, iterations=5, render=render)
|
||||
|
||||
# Altered basis functions
|
||||
obs1 = example_custom_mp("Reacher5dProMP-v0", seed=10, iterations=5, render=render)
|
||||
|
||||
# Custom MP
|
||||
example_fully_custom_mp(seed=10, iterations=1, render=render)
|
||||
@@ -0,0 +1,37 @@
|
||||
import fancy_gym
|
||||
|
||||
|
||||
def example_mp(env_name, seed=1, render=True):
|
||||
"""
|
||||
Example for running a movement primitive based version of a OpenAI-gym environment, which is already registered.
|
||||
For more information on movement primitive specific stuff, look at the traj_gen examples.
|
||||
Args:
|
||||
env_name: ProMP env_id
|
||||
seed: seed
|
||||
render: boolean
|
||||
Returns:
|
||||
|
||||
"""
|
||||
# While in this case gym.make() is possible to use as well, we recommend our custom make env function.
|
||||
env = fancy_gym.make(env_name, seed)
|
||||
|
||||
returns = 0
|
||||
obs = env.reset()
|
||||
# number of samples/full trajectories (multiple environment steps)
|
||||
for i in range(10):
|
||||
if render and i % 2 == 0:
|
||||
env.render(mode="human")
|
||||
else:
|
||||
env.render(mode=None)
|
||||
ac = env.action_space.sample()
|
||||
obs, reward, done, info = env.step(ac)
|
||||
returns += reward
|
||||
|
||||
if done:
|
||||
print(returns)
|
||||
obs = env.reset()
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
example_mp("ReacherProMP-v2")
|
||||
|
||||
@@ -0,0 +1,69 @@
|
||||
from collections import OrderedDict
|
||||
|
||||
import numpy as np
|
||||
from matplotlib import pyplot as plt
|
||||
|
||||
import fancy_gym
|
||||
|
||||
# This might work for some environments, however, please verify either way the correct trajectory information
|
||||
# for your environment are extracted below
|
||||
SEED = 1
|
||||
|
||||
env_id = "Reacher5dProMP-v0"
|
||||
|
||||
env = fancy_gym.make(env_id, seed=SEED, controller_kwargs={'p_gains': 0.05, 'd_gains': 0.05}).env
|
||||
env.action_space.seed(SEED)
|
||||
|
||||
# Plot difference between real trajectory and target MP trajectory
|
||||
env.reset()
|
||||
w = env.action_space.sample()
|
||||
pos, vel = env.get_trajectory(w)
|
||||
|
||||
base_shape = env.env.action_space.shape
|
||||
actual_pos = np.zeros((len(pos), *base_shape))
|
||||
actual_vel = np.zeros((len(pos), *base_shape))
|
||||
act = np.zeros((len(pos), *base_shape))
|
||||
|
||||
plt.ion()
|
||||
fig = plt.figure()
|
||||
ax = fig.add_subplot(1, 1, 1)
|
||||
|
||||
img = ax.imshow(env.env.render(mode="rgb_array"))
|
||||
fig.show()
|
||||
|
||||
for t, pos_vel in enumerate(zip(pos, vel)):
|
||||
actions = env.tracking_controller.get_action(pos_vel[0], pos_vel[1], env.current_vel, env.current_pos)
|
||||
actions = np.clip(actions, env.env.action_space.low, env.env.action_space.high)
|
||||
_, _, _, _ = env.env.step(actions)
|
||||
if t % 15 == 0:
|
||||
img.set_data(env.env.render(mode="rgb_array"))
|
||||
fig.canvas.draw()
|
||||
fig.canvas.flush_events()
|
||||
act[t, :] = actions
|
||||
# TODO verify for your environment
|
||||
actual_pos[t, :] = env.current_pos
|
||||
actual_vel[t, :] = env.current_vel
|
||||
|
||||
plt.figure(figsize=(15, 5))
|
||||
|
||||
plt.subplot(131)
|
||||
plt.title("Position")
|
||||
p1 = plt.plot(actual_pos, c='C0', label="true")
|
||||
p2 = plt.plot(pos, c='C1', label="MP")
|
||||
plt.xlabel("Episode steps")
|
||||
handles, labels = plt.gca().get_legend_handles_labels()
|
||||
|
||||
by_label = OrderedDict(zip(labels, handles))
|
||||
plt.legend(by_label.values(), by_label.keys())
|
||||
|
||||
plt.subplot(132)
|
||||
plt.title("Velocity")
|
||||
plt.plot(actual_vel, c='C0', label="true")
|
||||
plt.plot(vel, c='C1', label="MP")
|
||||
plt.xlabel("Episode steps")
|
||||
|
||||
plt.subplot(133)
|
||||
plt.title(f"Actions {np.std(act, axis=0)}")
|
||||
plt.plot(act, c="C0"),
|
||||
plt.xlabel("Episode steps")
|
||||
plt.show()
|
||||
Reference in New Issue
Block a user