Donkey Car OpenAI Gym
- Free software: MIT license
- Documentation: https://gym-donkeycar.readthedocs.io/en/latest/
- Install with pip
pip install gym-donkeycar
- Download simulator binaries: https://github.com/tawnkramer/gym-donkeycar/releases
A short and compact introduction for people who know gym environments, but want to understand this one. Simple example code:
import os
import gym
import gym_donkeycar
import numpy as np
#%% SET UP ENVIRONMENT
os.environ['DONKEY_SIM_PATH'] = f"{PATH_TO_APP}/donkey_sim.app/Contents/MacOS/donkey_sim"
os.environ['DONKEY_SIM_PORT'] = str(9091)
os.environ['DONKEY_SIM_HEADLESS'] = str(0) # "1" is headless
env = gym.make("donkey-warehouse-v0")
#%% PLAY
obv = env.reset()
for t in range(100):
action = np.array([0.0,0.5]) # drive straight with small speed
# execute the action
obv, reward, done, info = env.step(action)
- see more examples: https://github.com/tawnkramer/gym-donkeycar/tree/master/examples
A permissable action is a numpy array of length two with first steering and throttle, respectively. E.g. np.array([0,1]) goes straight at full speed, np.array([-5,1]) turns left etc.
Action Space: Box(2,)
Action names: ['steer', 'throttle']
- obv: The image that the donkey is seeing (np.array shape (120,160,3))
- reward: a reward that combines game over, how far from center and speed (max=1, min approx -2)
- done: Boolean. Game over if cte > max_cte or hit != "none"
- info contains: - cte: Cross track error (how far from center line) - positions: x,y,z - speed: positive forward, negative backward - hit: 'none' if all is good.
example info:
{'pos': (51.49209, 0.7399381, 117.3004),
'cte': -5.865292,
'speed': 9.319956,
'hit': 'none'}
- "donkey-warehouse-v0"
- "donkey-generated-roads-v0"
- "donkey-avc-sparkfun-v0"
- "donkey-generated-track-v0"
Original Source Code
Tawn Kramer - https://github.com/tawnkramer/gym-donkeycar
Roma Sokolkov - https://github.com/r7vme/gym-donkeycar cloned with permission from https://github.com/tawnkramer/sdsandbox
Release Engineer
This package was created with Cookiecutter and the audreyr/cookiecutter-pypackage project template.