search

aidudezzz / deepbots

A wrapper framework for Reinforcement Learning in the Webots robot simulator using Python 3.
https://deepbots.readthedocs.io/
GNU General Public License v3.0
230 stars 49 forks source link

Gym integration and general improvements #53

Closed tsampazk closed 3 years ago

tsampazk commented 3 years ago

Major refactoring and improvements in docstrings

List of main changes:

  1. SupervisorEnv now inherits both from gym.Env and Webots controller.Supervisor. All calls to self.supervisor... changed to self...
  2. Removed all ABC stuff for abstract methods, now every abstract method just raises a NotImplementedError, similar to gym
  3. Docstring additions and improvements
  4. Changed imports that used from .a_deepbots_package to from deepbots.a_deepbots_package
  5. Default reset implementation now steps the controller once
  6. Added gym in requirements.txt

~I added a TODO comment in all get_timestep methods. This refers to changing timestep to a pythonic class property, we can discuss this further.~

tsampazk commented 3 years ago

Renamed timestep, added @property getter/setter for timestep and added a deprecation warning in get_timestep.

Also fixed the emitter/receiver name initialization argument issue.