search

robotology / gym-ignition

Framework for developing OpenAI Gym robotics environments simulated with Ignition Gazebo
https://robotology.github.io/gym-ignition
GNU Lesser General Public License v3.0
227 stars 26 forks source link

TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases #428

Open psh9002 opened 2 years ago

psh9002 commented 2 years ago

Description:

There is a TypeError when I type 'python panda_pick_and_place.py'. I tried both Stable and Nightly.

Steps to reproduce

  1. Install Ignition Gazebo (Fortress) from source https://gazebosim.org/docs/fortress/install_ubuntu_src
  2. Install gym-ignition from https://robotology.github.io/gym-ignition/master/installation/nightly.html

Additional context

TypeError TypeError_anaconda

Traceback (most recent call last):
  File "panda_pick_and_place.py", line 7, in <module>
    import gym_ignition_environments
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition_environments/__init__.py", line 8, in <module>
    from . import models, randomizers, tasks
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition_environments/randomizers/__init__.py", line 5, in <module>
    from . import cartpole, cartpole_no_rand
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition_environments/randomizers/cartpole.py", line 8, in <module>
    from gym_ignition import randomizers, utils
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition/randomizers/__init__.py", line 5, in <module>
    from . import abc, gazebo_env_randomizer, model, physics
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition/randomizers/abc.py", line 7, in <module>
    import gym_ignition.base.task
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition/base/__init__.py", line 6, in <module>
    from . import runtime, task
  File "/home/bak/venv/lib/python3.8/site-packages/gym_ignition/base/runtime.py", line 11, in <module>
    class Runtime(gym.Env, abc.ABC):
TypeError: metaclass conflict: the metaclass of a derived class must be a (non-strict) subclass of the metaclasses of all its bases

Environment

psh9002 commented 2 years ago

I fix the problem by modifying runtime.py and gazebo_env_randomizer.py.

In runtime.py, I changed class Runtime(gym.Env, abc.ABC): to class Runtime(gym.Env, abc.ABCMeta): and, added @abc.abstractmethod above def timestamp(self) -> float:.

In gazebo_env_randomizer.py, I added class GazeboEnvRandomizer_Meta(type(gym.Wrapper), type(TaskRandomizer)): pass and changed class GazeboEnvRandomizer(gym.Wrapper, TaskRandomizer, abc.ABC): to class GazeboEnvRandomizer(gym.Wrapper, TaskRandomizer, metaclass=GazeboEnvRandomizer_Meta):

psh9002 commented 2 years ago

However, there is another issue.

[Err] [Conversions.cc:1757] Tried to convert SDF [world] into [plugin]
[Err] [Joint.cpp:756] The active joint control mode does not accept a velocity target
[Err] [Model.cpp:1259] Failed to set force of joint 'panda_joint1'
Traceback (most recent call last):
  File "/home/bak/Projects/ignition-robot/examples/manipulation.py", line 268, in <module>
    add_panda_controller(panda=panda, controller_period=gazebo.step_size())
  File "/home/bak/Projects/ignition-robot/examples/manipulation.py", line 52, in add_panda_controller
    assert panda.set_joint_velocity_targets(panda.joint_velocities())
AssertionError

Gazebo Gui window was opened but, there is only empty space. It is also the same on the pendulum example at ScenarIO.

empty_gui

giu950 commented 1 year ago

Hello @psh9002, I'm having the same issue, do you have any tips to solve it?

psh9002 commented 1 year ago

Hello @giu950 , Unfortunately no. Now, I'm not using Gazebo

DanielePucci commented 1 year ago

CC @diegoferigo

diegoferigo commented 1 year ago

Hi everyone, thanks for the feedback. I described the current status of the project in https://github.com/robotology/gym-ignition/issues/430 in case someone wants to start addressing the major problems affecting the current status of the project.

Unfortunately there have been major changes of our main dependencies (both Ignition Gazebo and OpenAI Gym), and my research focus on the past 2 years shifted towards solutions providing parallelization on hardware accelerators like GPUs and TPUs. I'm not sure if there is anyone in AMI interested in the short term to pick up what's left, for sure in the close future I don't think I will find enough bandwidth for a proper upstream alignment, but I'd be happy to provide all the needed support and guidance.