Gymnasium-Planar-Robotics#

Gymnasium-Planar-Robotics (GymPR) is a library that contains reinforcement learning (RL) environments for motion planning and object manipulation in the field of planar robotics. The main component of every environment is a magnetic levitation system, a so-called planar motor system, that consists of two basic components that are shown in the Figure below. Firstly, mobile permanent magnets, so-called movers, consist of a housing and a complex permanent magnet structure on the lower side of the mover. Secondly, static motor modules, so-called tiles, are the active components of the drive system. As shown in the Figure, the tiles enable the coil-induced emission of electromagnetic fields (yellow) that interact with the mover’s field (blue). During operation, the movers hover above the tiles and can be controlled in six dimensions by adjusting the currents in the coils contained in the tiles.

_images/visual_abstract.png


The six degrees of freedom of a mover are shown in the following picture:

_images/img_mover_dofs.png


GymPR is based on the MuJoCo physics engine and all environments follow either the Gymnasium API for single-agent RL or the PettingZoo parallel API for multi-agent RL. The following example shows how to use a trained policy with an example environment that follows the Gymnasium API:

import gymnasium as gym

env = gym.make("BenchmarkPushingEnv-v0", render_mode="human")
observation, info = env.reset(seed=42)

for _ in range(0,100):
   while not terminated and not truncated:
      action = policy(observation)  # custom policy
      observation, reward, terminated, truncated, info = env.step(action)

   observation, info = env.reset()
env.close()

Due to the use of standard RL APIs, GymPR is compatible with common RL libraries, such as Stable-Baselines3 or Tianshou. The library contains a collection of existing RL environments in the field of planar robotics, which will continue to be updated in the future, as well as basic single-agent and multi-agent environments that serve as starting points for the easy development of new research-specific environments.

Contents: