def test_real_env_contructors(dt, max_steps):
qbbr = QBallBalancerReal(dt=dt, max_steps=max_steps)
assert qbbr is not None
qcp_st = QCartPoleStabReal(dt=dt, max_steps=max_steps)
assert qcp_st is not None
qcp_su = QCartPoleSwingUpReal(dt=dt, max_steps=max_steps)
assert qcp_su is not None
qqr = QQubeReal(dt=dt, max_steps=max_steps)
assert qqr is not None
if __name__ == '__main__':
# Parse command line arguments
args = get_argparser().parse_args()
# Set up PD-controller
if args.env_name in QBallBalancerReal.name:
env = QBallBalancerReal(args.dt, args.max_steps)
policy = QBallBalancerPDCtrl(env.spec,
kp=to.diag(to.tensor([3.45, 3.45])),
kd=to.diag(to.tensor([2.11, 2.11])))
print_cbt('Set up controller for the QBallBalancerReal environment.',
'c')
elif args.env_name == QCartPoleStabReal.name:
env = QCartPoleStabReal(args.dt, args.max_steps)
policy = CartpoleStabilizerPolicy(env.spec,
K=np.array([
1.2278416e+00, 4.5279346e+00,
-1.2385756e-02, 6.0038762e+00,
-4.1818547e+00
]))
# policy = QCartPoleSwingUpAndBalanceCtrl(env.spec)
print_cbt('Set up controller for the QCartPoleStabReal environment.',
'c')
elif args.env_name == QCartPoleSwingUpReal.name:
env = QCartPoleSwingUpReal(args.dt, args.max_steps)
policy = QCartPoleSwingUpAndBalanceCtrl(env.spec)
print_cbt(
'Set up controller for the QCartPoleSwingUpReal environment.', 'c')
if __name__ == '__main__':
# Parse command line arguments
args = get_argparser().parse_args()
# Get the experiment's directory to load from
ex_dir = ask_for_experiment()
ex_tag = ex_dir.split('--', 1)[1]
# Load the policy and the environment (for constructing the real-world counterpart)
env_sim, policy, _ = load_experiment(ex_dir)
if args.verbose:
print(f'Policy params:\n{policy.param_values.detach().numpy()}')
# Create real-world counterpart (without domain randomization)
env_real = QCartPoleStabReal(args.dt, args.max_steps)
print_cbt('Set up the QCartPoleStabReal environment.', 'c')
# Set up the disturber
disturber_pos = TimePolicy(env_real.spec, volt_disturbance_pos,
env_real.dt)
disturber_neg = TimePolicy(env_real.spec, volt_disturbance_neg,
env_real.dt)
steps_disturb = 10
print_cbt(
f'Set up the disturbers for the QCartPoleStabReal environment.'
f'\nVolt disturbance: {6} volts for {steps_disturb} steps', 'c')
# Center cart and reset velocity filters and wait until the user or the conroller has put pole upright
env_real.reset()
print_cbt('Ready', 'g')
def default_qcpst_real():
return QCartPoleStabReal(dt=1 / 500.0, max_steps=500)