Gym environment for Bicycle model
For RL zoo: clone this repo into the RL-zoo folder from
https://github.com/DLR-RM/rl-baselines3-zoo
Add the following into hyperparams/ppo.yml
# custom
BicycleKin-v0:
vec_env_wrapper:
- stable_baselines3.common.vec_env.VecMonitor
- stable_baselines3.common.vec_env.vec_normalize.VecNormalize
n_envs: 20
n_timesteps: !!float 1e7
policy: 'MlpPolicy'
n_steps: 20_000
gae_lambda: 0.95
gamma: 0.9
n_epochs: 10
ent_coef: 0.0
learning_rate: !!float 1e-3
clip_range: 0.2
use_sde: True
sde_sample_freq: 4
Also add the following to utils/import_envs.py
register(id="BicycleKin-v0", entry_point="gym_bicycle.envs:BicycleKin")
or
gym.envs.register( id='BicycleKin-v0', entry_point='gym_bicycle.envs:BicycleKin')
try running
python train.py --algo ppo --env "BicycleKin-v0" --vec-env "subproc"