initial commit

scripts/play.py

@@ -0,0 +1,162 @@
+import isaacgym
+
+assert isaacgym
+import torch
+import numpy as np
+
+import glob
+import pickle as pkl
+
+from go1_gym.envs import *
+from go1_gym.envs.base.legged_robot_config import Cfg
+from go1_gym.envs.go1.go1_config import config_go1
+from go1_gym.envs.go1.velocity_tracking import VelocityTrackingEasyEnv
+
+from tqdm import tqdm
+
+def load_policy(logdir):
+    body = torch.jit.load(logdir + '/checkpoints/body_latest.jit')
+    import os
+    adaptation_module = torch.jit.load(logdir + '/checkpoints/adaptation_module_latest.jit')
+
+    def policy(obs, info={}):
+        i = 0
+        latent = adaptation_module.forward(obs["obs_history"].to('cpu'))
+        action = body.forward(torch.cat((obs["obs_history"].to('cpu'), latent), dim=-1))
+        info['latent'] = latent
+        return action
+
+    return policy
+
+
+def load_env(label, headless=False):
+    dirs = glob.glob(f"../runs/{label}/*")
+    logdir = sorted(dirs)[0]
+
+    with open(logdir + "/parameters.pkl", 'rb') as file:
+        pkl_cfg = pkl.load(file)
+        print(pkl_cfg.keys())
+        cfg = pkl_cfg["Cfg"]
+        print(cfg.keys())
+
+        for key, value in cfg.items():
+            if hasattr(Cfg, key):
+                for key2, value2 in cfg[key].items():
+                    setattr(getattr(Cfg, key), key2, value2)
+
+    # turn off DR for evaluation script
+    Cfg.domain_rand.push_robots = False
+    Cfg.domain_rand.randomize_friction = False
+    Cfg.domain_rand.randomize_gravity = False
+    Cfg.domain_rand.randomize_restitution = False
+    Cfg.domain_rand.randomize_motor_offset = False
+    Cfg.domain_rand.randomize_motor_strength = False
+    Cfg.domain_rand.randomize_friction_indep = False
+    Cfg.domain_rand.randomize_ground_friction = False
+    Cfg.domain_rand.randomize_base_mass = False
+    Cfg.domain_rand.randomize_Kd_factor = False
+    Cfg.domain_rand.randomize_Kp_factor = False
+    Cfg.domain_rand.randomize_joint_friction = False
+    Cfg.domain_rand.randomize_com_displacement = False
+
+    Cfg.env.num_recording_envs = 1
+    Cfg.env.num_envs = 1
+    Cfg.terrain.num_rows = 5
+    Cfg.terrain.num_cols = 5
+    Cfg.terrain.border_size = 0
+    Cfg.terrain.center_robots = True
+    Cfg.terrain.center_span = 1
+    Cfg.terrain.teleport_robots = True
+
+    Cfg.domain_rand.lag_timesteps = 6
+    Cfg.domain_rand.randomize_lag_timesteps = True
+    Cfg.control.control_type = "actuator_net"
+
+    from go1_gym.envs.wrappers.history_wrapper import HistoryWrapper
+
+    env = VelocityTrackingEasyEnv(sim_device='cuda:0', headless=False, cfg=Cfg)
+    env = HistoryWrapper(env)
+
+    # load policy
+    from ml_logger import logger
+    from go1_gym_learn.ppo_cse.actor_critic import ActorCritic
+
+    policy = load_policy(logdir)
+
+    return env, policy
+
+
+def play_go1(headless=True):
+    from ml_logger import logger
+
+    from pathlib import Path
+    from go1_gym import MINI_GYM_ROOT_DIR
+    import glob
+    import os
+
+    label = "gait-conditioned-agility/pretrain-v0/train"
+
+    env, policy = load_env(label, headless=headless)
+
+    num_eval_steps = 250
+    gaits = {"pronking": [0, 0, 0],
+             "trotting": [0.5, 0, 0],
+             "bounding": [0, 0.5, 0],
+             "pacing": [0, 0, 0.5]}
+
+    x_vel_cmd, y_vel_cmd, yaw_vel_cmd = 1.5, 0.0, 0.0
+    body_height_cmd = 0.0
+    step_frequency_cmd = 3.0
+    gait = torch.tensor(gaits["trotting"])
+    footswing_height_cmd = 0.08
+    pitch_cmd = 0.0
+    roll_cmd = 0.0
+    stance_width_cmd = 0.25
+
+    measured_x_vels = np.zeros(num_eval_steps)
+    target_x_vels = np.ones(num_eval_steps) * x_vel_cmd
+    joint_positions = np.zeros((num_eval_steps, 12))
+
+    obs = env.reset()
+
+    for i in tqdm(range(num_eval_steps)):
+        with torch.no_grad():
+            actions = policy(obs)
+        env.commands[:, 0] = x_vel_cmd
+        env.commands[:, 1] = y_vel_cmd
+        env.commands[:, 2] = yaw_vel_cmd
+        env.commands[:, 3] = body_height_cmd
+        env.commands[:, 4] = step_frequency_cmd
+        env.commands[:, 5:8] = gait
+        env.commands[:, 8] = 0.5
+        env.commands[:, 9] = footswing_height_cmd
+        env.commands[:, 10] = pitch_cmd
+        env.commands[:, 11] = roll_cmd
+        env.commands[:, 12] = stance_width_cmd
+        obs, rew, done, info = env.step(actions)
+
+        measured_x_vels[i] = env.base_lin_vel[0, 0]
+        joint_positions[i] = env.dof_pos[0, :].cpu()
+
+    # plot target and measured forward velocity
+    from matplotlib import pyplot as plt
+    fig, axs = plt.subplots(2, 1, figsize=(12, 5))
+    axs[0].plot(np.linspace(0, num_eval_steps * env.dt, num_eval_steps), measured_x_vels, color='black', linestyle="-", label="Measured")
+    axs[0].plot(np.linspace(0, num_eval_steps * env.dt, num_eval_steps), target_x_vels, color='black', linestyle="--", label="Desired")
+    axs[0].legend()
+    axs[0].set_title("Forward Linear Velocity")
+    axs[0].set_xlabel("Time (s)")
+    axs[0].set_ylabel("Velocity (m/s)")
+
+    axs[1].plot(np.linspace(0, num_eval_steps * env.dt, num_eval_steps), joint_positions, linestyle="-", label="Measured")
+    axs[1].set_title("Joint Positions")
+    axs[1].set_xlabel("Time (s)")
+    axs[1].set_ylabel("Joint Position (rad)")
+
+    plt.tight_layout()
+    plt.show()
+
+
+if __name__ == '__main__':
+    # to see the environment rendering, set headless=False
+    play_go1(headless=False)

scripts/test.py

@@ -0,0 +1,206 @@
+import isaacgym
+
+assert isaacgym
+import matplotlib.pyplot as plt
+import torch
+from tqdm import trange
+
+from go1_gym.envs import *
+from go1_gym.envs.base.legged_robot_config import Cfg
+from go1_gym.envs.go1.go1_config import config_go1
+from go1_gym.envs.go1.velocity_tracking import VelocityTrackingEasyEnv
+
+
+def run_env(render=False, headless=False):
+    # prepare environment
+    config_go1(Cfg)
+
+    Cfg.commands.num_lin_vel_bins = 30
+    Cfg.commands.num_ang_vel_bins = 30
+    Cfg.curriculum_thresholds.tracking_ang_vel = 0.7
+    Cfg.curriculum_thresholds.tracking_lin_vel = 0.8
+    Cfg.curriculum_thresholds.tracking_contacts_shaped_vel = 0.9
+    Cfg.curriculum_thresholds.tracking_contacts_shaped_force = 0.9
+
+    Cfg.commands.distributional_commands = True
+
+    Cfg.env.priv_observe_motion = False
+    Cfg.env.priv_observe_gravity_transformed_motion = True
+    Cfg.domain_rand.randomize_friction_indep = False
+    Cfg.env.priv_observe_friction_indep = False
+    Cfg.domain_rand.randomize_friction = True
+    Cfg.env.priv_observe_friction = False
+    Cfg.domain_rand.friction_range = [0.0, 0.0]
+    Cfg.domain_rand.randomize_restitution = True
+    Cfg.env.priv_observe_restitution = False
+    Cfg.domain_rand.restitution_range = [0.0, 1.0]
+    Cfg.domain_rand.randomize_base_mass = True
+    Cfg.env.priv_observe_base_mass = False
+    Cfg.domain_rand.added_mass_range = [-1.0, 3.0]
+    Cfg.domain_rand.randomize_gravity = True
+    Cfg.domain_rand.gravity_range = [-2.0, 2.0]
+    Cfg.domain_rand.gravity_rand_interval_s = 2.0
+    Cfg.domain_rand.gravity_impulse_duration = 0.5
+    Cfg.env.priv_observe_gravity = True
+    Cfg.domain_rand.randomize_com_displacement = False
+    Cfg.domain_rand.com_displacement_range = [-0.15, 0.15]
+    Cfg.env.priv_observe_com_displacement = False
+    Cfg.domain_rand.randomize_ground_friction = True
+    Cfg.env.priv_observe_ground_friction = False
+    Cfg.env.priv_observe_ground_friction_per_foot = False
+    Cfg.domain_rand.ground_friction_range = [0.3, 2.0]
+    Cfg.domain_rand.randomize_motor_strength = True
+    Cfg.domain_rand.motor_strength_range = [0.9, 1.1]
+    Cfg.env.priv_observe_motor_strength = False
+    Cfg.domain_rand.randomize_motor_offset = True
+    Cfg.domain_rand.motor_offset_range = [-0.02, 0.02]
+    Cfg.env.priv_observe_motor_offset = False
+    Cfg.domain_rand.push_robots = False
+    Cfg.domain_rand.randomize_Kp_factor = False
+    Cfg.env.priv_observe_Kp_factor = False
+    Cfg.domain_rand.randomize_Kd_factor = False
+    Cfg.env.priv_observe_Kd_factor = False
+    Cfg.env.priv_observe_body_velocity = False
+    Cfg.env.priv_observe_body_height = False
+    Cfg.env.priv_observe_desired_contact_states = False
+    Cfg.env.priv_observe_contact_forces = False
+    Cfg.env.priv_observe_foot_displacement = False
+
+    Cfg.env.num_privileged_obs = 3
+    Cfg.env.num_observation_history = 30
+    Cfg.reward_scales.feet_contact_forces = 0.0
+
+    Cfg.domain_rand.rand_interval_s = 4
+    Cfg.commands.num_commands = 15
+    Cfg.env.observe_two_prev_actions = True
+    Cfg.env.observe_yaw = True
+    Cfg.env.num_observations = 71
+    Cfg.env.num_scalar_observations = 71
+    Cfg.env.observe_gait_commands = True
+    Cfg.env.observe_timing_parameter = False
+    Cfg.env.observe_clock_inputs = True
+
+    Cfg.domain_rand.tile_height_range = [-0.0, 0.0]
+    Cfg.domain_rand.tile_height_curriculum = False
+    Cfg.domain_rand.tile_height_update_interval = 1000, 3000
+    Cfg.domain_rand.tile_height_curriculum_step = 0.01
+    Cfg.terrain.border_size = 0.0
+
+    Cfg.commands.resampling_time = 10
+
+    Cfg.reward_scales.feet_slip = -0.04
+    Cfg.reward_scales.action_smoothness_1 = -0.1
+    Cfg.reward_scales.action_smoothness_2 = -0.1
+    Cfg.reward_scales.dof_vel = -1e-4
+    Cfg.reward_scales.dof_pos = -0.05
+    Cfg.reward_scales.jump = 10.0
+    Cfg.reward_scales.base_height = 0.0
+    Cfg.rewards.base_height_target = 0.30
+    Cfg.reward_scales.estimation_bonus = 0.0
+
+    Cfg.reward_scales.feet_impact_vel = -0.0
+
+    # rewards.footswing_height = 0.09
+    Cfg.reward_scales.feet_clearance = -0.0
+    Cfg.reward_scales.feet_clearance_cmd = -15.
+
+    # reward_scales.feet_contact_forces = -0.01
+
+    Cfg.rewards.reward_container_name = "CoRLRewards"
+    Cfg.rewards.only_positive_rewards = False
+    Cfg.rewards.only_positive_rewards_ji22_style = True
+    Cfg.rewards.sigma_rew_neg = 0.02
+
+    Cfg.reward_scales.hop_symmetry = 0.0
+    Cfg.rewards.kappa_gait_probs = 0.07
+    Cfg.rewards.gait_force_sigma = 100.
+    Cfg.rewards.gait_vel_sigma = 10.
+
+    Cfg.reward_scales.tracking_contacts_shaped_force = 4.0
+    Cfg.reward_scales.tracking_contacts_shaped_vel = 4.0
+
+    Cfg.reward_scales.collision = -5.0
+
+    Cfg.commands.lin_vel_x = [-1.0, 1.0]
+    Cfg.commands.lin_vel_y = [-0.6, 0.6]
+    Cfg.commands.ang_vel_yaw = [-1.0, 1.0]
+    Cfg.commands.body_height_cmd = [-0.25, 0.15]
+    Cfg.commands.gait_frequency_cmd_range = [1.5, 4.0]
+    Cfg.commands.gait_phase_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_offset_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_bound_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_duration_cmd_range = [0.5, 0.5]
+    Cfg.commands.footswing_height_range = [0.03, 0.25]
+
+    Cfg.reward_scales.lin_vel_z = -0.02
+    Cfg.reward_scales.ang_vel_xy = -0.001
+    Cfg.reward_scales.base_height = 0.0
+    Cfg.reward_scales.feet_air_time = 0.0
+
+    Cfg.commands.limit_vel_x = [-5.0, 5.0]
+    Cfg.commands.limit_vel_y = [-0.6, 0.6]
+    Cfg.commands.limit_vel_yaw = [-5.0, 5.0]
+    Cfg.commands.limit_body_height = [-0.25, 0.15]
+    Cfg.commands.limit_gait_frequency = [1.5, 4.0]
+    Cfg.commands.limit_gait_phase = [0.0, 1.0]
+    Cfg.commands.limit_gait_offset = [0.0, 1.0]
+    Cfg.commands.limit_gait_bound = [0.0, 1.0]
+    Cfg.commands.limit_gait_duration = [0.5, 0.5]
+    Cfg.commands.limit_footswing_height = [0.03, 0.25]
+
+    Cfg.commands.num_bins_vel_x = 21
+    Cfg.commands.num_bins_vel_y = 1
+    Cfg.commands.num_bins_vel_yaw = 21
+    Cfg.commands.num_bins_body_height = 1
+    Cfg.commands.num_bins_gait_frequency = 1
+    Cfg.commands.num_bins_gait_phase = 1
+    Cfg.commands.num_bins_gait_offset = 1
+    Cfg.commands.num_bins_gait_bound = 1
+    Cfg.commands.num_bins_gait_duration = 1
+    Cfg.commands.num_bins_footswing_height = 1
+
+    Cfg.normalization.friction_range = [0, 1]
+    Cfg.normalization.ground_friction_range = [0, 1]
+
+    Cfg.commands.exclusive_phase_offset = False
+    Cfg.commands.pacing_offset = False
+    Cfg.commands.binary_phases = True
+    Cfg.commands.gaitwise_curricula = True
+
+    # 5 times per second
+
+    Cfg.env.num_envs = 3
+    Cfg.domain_rand.push_interval_s = 1
+    Cfg.terrain.num_rows = 3
+    Cfg.terrain.num_cols = 5
+    Cfg.terrain.border_size = 0
+    Cfg.domain_rand.randomize_friction = True
+    Cfg.domain_rand.friction_range = [1.0, 1.01]
+    Cfg.domain_rand.randomize_base_mass = True
+    Cfg.domain_rand.added_mass_range = [0., 6.]
+    Cfg.terrain.terrain_noise_magnitude = 0.0
+    # Cfg.asset.fix_base_link = True
+
+    Cfg.domain_rand.lag_timesteps = 6
+    Cfg.domain_rand.randomize_lag_timesteps = True
+    Cfg.control.control_type = "actuator_net"
+
+    env = VelocityTrackingEasyEnv(sim_device='cuda:0', headless=False, cfg=Cfg)
+    env.reset()
+
+    if render and headless:
+        img = env.render(mode="rgb_array")
+        plt.imshow(img)
+        plt.show()
+        print("Show the first frame and exit.")
+        exit()
+
+    for i in trange(1000, desc="Running"):
+        actions = 0. * torch.ones(env.num_envs, env.num_actions, device=env.device)
+        obs, rew, done, info = env.step(actions)
+
+    print("Done")
+
+
+if __name__ == '__main__':
+    run_env(render=True, headless=False)

scripts/train.py

@@ -0,0 +1,256 @@
+def train_go1(headless=True):
+
+    import isaacgym
+    assert isaacgym
+    import torch
+
+    from go1_gym.envs.base.legged_robot_config import Cfg
+    from go1_gym.envs.go1.go1_config import config_go1
+    from go1_gym.envs.go1.velocity_tracking import VelocityTrackingEasyEnv
+
+    from ml_logger import logger
+
+    from go1_gym_learn.ppo_cse import Runner
+    from go1_gym.envs.wrappers.history_wrapper import HistoryWrapper
+    from go1_gym_learn.ppo_cse.actor_critic import AC_Args
+    from go1_gym_learn.ppo_cse.ppo import PPO_Args
+    from go1_gym_learn.ppo_cse import RunnerArgs
+
+    config_go1(Cfg)
+
+    Cfg.commands.num_lin_vel_bins = 30
+    Cfg.commands.num_ang_vel_bins = 30
+    Cfg.curriculum_thresholds.tracking_ang_vel = 0.7
+    Cfg.curriculum_thresholds.tracking_lin_vel = 0.8
+    Cfg.curriculum_thresholds.tracking_contacts_shaped_vel = 0.90
+    Cfg.curriculum_thresholds.tracking_contacts_shaped_force = 0.90
+
+    Cfg.commands.distributional_commands = True
+
+    Cfg.domain_rand.lag_timesteps = 6
+    Cfg.domain_rand.randomize_lag_timesteps = True
+    Cfg.control.control_type = "actuator_net"
+
+    Cfg.domain_rand.randomize_rigids_after_start = False
+    Cfg.env.priv_observe_motion = False
+    Cfg.env.priv_observe_gravity_transformed_motion = False
+    Cfg.domain_rand.randomize_friction_indep = False
+    Cfg.env.priv_observe_friction_indep = False
+    Cfg.domain_rand.randomize_friction = True
+    Cfg.env.priv_observe_friction = True
+    Cfg.domain_rand.friction_range = [0.1, 3.0]
+    Cfg.domain_rand.randomize_restitution = True
+    Cfg.env.priv_observe_restitution = True
+    Cfg.domain_rand.restitution_range = [0.0, 0.4]
+    Cfg.domain_rand.randomize_base_mass = True
+    Cfg.env.priv_observe_base_mass = False
+    Cfg.domain_rand.added_mass_range = [-1.0, 3.0]
+    Cfg.domain_rand.randomize_gravity = True
+    Cfg.domain_rand.gravity_range = [-2.0, 2.0]
+    Cfg.domain_rand.gravity_rand_interval_s = 8.0
+    Cfg.domain_rand.gravity_impulse_duration = 0.99
+    Cfg.env.priv_observe_gravity = False
+    Cfg.domain_rand.randomize_com_displacement = False
+    Cfg.domain_rand.com_displacement_range = [-0.15, 0.15]
+    Cfg.env.priv_observe_com_displacement = False
+    Cfg.domain_rand.randomize_ground_friction = True
+    Cfg.env.priv_observe_ground_friction = False
+    Cfg.env.priv_observe_ground_friction_per_foot = False
+    Cfg.domain_rand.ground_friction_range = [0.0, 0.0]
+    Cfg.domain_rand.randomize_motor_strength = True
+    Cfg.domain_rand.motor_strength_range = [0.9, 1.1]
+    Cfg.env.priv_observe_motor_strength = False
+    Cfg.domain_rand.randomize_motor_offset = True
+    Cfg.domain_rand.motor_offset_range = [-0.02, 0.02]
+    Cfg.env.priv_observe_motor_offset = False
+    Cfg.domain_rand.push_robots = False
+    Cfg.domain_rand.randomize_Kp_factor = False
+    Cfg.env.priv_observe_Kp_factor = False
+    Cfg.domain_rand.randomize_Kd_factor = False
+    Cfg.env.priv_observe_Kd_factor = False
+    Cfg.env.priv_observe_body_velocity = False
+    Cfg.env.priv_observe_body_height = False
+    Cfg.env.priv_observe_desired_contact_states = False
+    Cfg.env.priv_observe_contact_forces = False
+    Cfg.env.priv_observe_foot_displacement = False
+    Cfg.env.priv_observe_gravity_transformed_foot_displacement = False
+
+    Cfg.env.num_privileged_obs = 2
+    Cfg.env.num_observation_history = 30
+    Cfg.reward_scales.feet_contact_forces = 0.0
+
+    Cfg.domain_rand.rand_interval_s = 4
+    Cfg.commands.num_commands = 15
+    Cfg.env.observe_two_prev_actions = True
+    Cfg.env.observe_yaw = False
+    Cfg.env.num_observations = 70
+    Cfg.env.num_scalar_observations = 70
+    Cfg.env.observe_gait_commands = True
+    Cfg.env.observe_timing_parameter = False
+    Cfg.env.observe_clock_inputs = True
+
+    Cfg.domain_rand.tile_height_range = [-0.0, 0.0]
+    Cfg.domain_rand.tile_height_curriculum = False
+    Cfg.domain_rand.tile_height_update_interval = 1000000
+    Cfg.domain_rand.tile_height_curriculum_step = 0.01
+    Cfg.terrain.border_size = 0.0
+    Cfg.terrain.mesh_type = "trimesh"
+    Cfg.terrain.num_cols = 30
+    Cfg.terrain.num_rows = 30
+    Cfg.terrain.terrain_width = 5.0
+    Cfg.terrain.terrain_length = 5.0
+    Cfg.terrain.x_init_range = 0.2
+    Cfg.terrain.y_init_range = 0.2
+    Cfg.terrain.teleport_thresh = 0.3
+    Cfg.terrain.teleport_robots = False
+    Cfg.terrain.center_robots = True
+    Cfg.terrain.center_span = 4
+    Cfg.terrain.horizontal_scale = 0.10
+    Cfg.rewards.use_terminal_foot_height = False
+    Cfg.rewards.use_terminal_body_height = True
+    Cfg.rewards.terminal_body_height = 0.05
+    Cfg.rewards.use_terminal_roll_pitch = True
+    Cfg.rewards.terminal_body_ori = 1.6
+
+    Cfg.commands.resampling_time = 10
+
+    Cfg.reward_scales.feet_slip = -0.04
+    Cfg.reward_scales.action_smoothness_1 = -0.1
+    Cfg.reward_scales.action_smoothness_2 = -0.1
+    Cfg.reward_scales.dof_vel = -1e-4
+    Cfg.reward_scales.dof_pos = -0.0
+    Cfg.reward_scales.jump = 10.0
+    Cfg.reward_scales.base_height = 0.0
+    Cfg.rewards.base_height_target = 0.30
+    Cfg.reward_scales.estimation_bonus = 0.0
+    Cfg.reward_scales.raibert_heuristic = -10.0
+    Cfg.reward_scales.feet_impact_vel = -0.0
+    Cfg.reward_scales.feet_clearance = -0.0
+    Cfg.reward_scales.feet_clearance_cmd = -0.0
+    Cfg.reward_scales.feet_clearance_cmd_linear = -30.0
+    Cfg.reward_scales.orientation = 0.0
+    Cfg.reward_scales.orientation_control = -5.0
+    Cfg.reward_scales.tracking_stance_width = -0.0
+    Cfg.reward_scales.tracking_stance_length = -0.0
+    Cfg.reward_scales.lin_vel_z = -0.02
+    Cfg.reward_scales.ang_vel_xy = -0.001
+    Cfg.reward_scales.feet_air_time = 0.0
+    Cfg.reward_scales.hop_symmetry = 0.0
+    Cfg.rewards.kappa_gait_probs = 0.07
+    Cfg.rewards.gait_force_sigma = 100.
+    Cfg.rewards.gait_vel_sigma = 10.
+    Cfg.reward_scales.tracking_contacts_shaped_force = 4.0
+    Cfg.reward_scales.tracking_contacts_shaped_vel = 4.0
+    Cfg.reward_scales.collision = -5.0
+
+    Cfg.rewards.reward_container_name = "CoRLRewards"
+    Cfg.rewards.only_positive_rewards = False
+    Cfg.rewards.only_positive_rewards_ji22_style = True
+    Cfg.rewards.sigma_rew_neg = 0.02
+
+
+
+    Cfg.commands.lin_vel_x = [-1.0, 1.0]
+    Cfg.commands.lin_vel_y = [-0.6, 0.6]
+    Cfg.commands.ang_vel_yaw = [-1.0, 1.0]
+    Cfg.commands.body_height_cmd = [-0.25, 0.15]
+    Cfg.commands.gait_frequency_cmd_range = [2.0, 4.0]
+    Cfg.commands.gait_phase_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_offset_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_bound_cmd_range = [0.0, 1.0]
+    Cfg.commands.gait_duration_cmd_range = [0.5, 0.5]
+    Cfg.commands.footswing_height_range = [0.03, 0.35]
+    Cfg.commands.body_pitch_range = [-0.4, 0.4]
+    Cfg.commands.body_roll_range = [-0.0, 0.0]
+    Cfg.commands.stance_width_range = [0.10, 0.45]
+    Cfg.commands.stance_length_range = [0.35, 0.45]
+
+    Cfg.commands.limit_vel_x = [-5.0, 5.0]
+    Cfg.commands.limit_vel_y = [-0.6, 0.6]
+    Cfg.commands.limit_vel_yaw = [-5.0, 5.0]
+    Cfg.commands.limit_body_height = [-0.25, 0.15]
+    Cfg.commands.limit_gait_frequency = [2.0, 4.0]
+    Cfg.commands.limit_gait_phase = [0.0, 1.0]
+    Cfg.commands.limit_gait_offset = [0.0, 1.0]
+    Cfg.commands.limit_gait_bound = [0.0, 1.0]
+    Cfg.commands.limit_gait_duration = [0.5, 0.5]
+    Cfg.commands.limit_footswing_height = [0.03, 0.35]
+    Cfg.commands.limit_body_pitch = [-0.4, 0.4]
+    Cfg.commands.limit_body_roll = [-0.0, 0.0]
+    Cfg.commands.limit_stance_width = [0.10, 0.45]
+    Cfg.commands.limit_stance_length = [0.35, 0.45]
+
+    Cfg.commands.num_bins_vel_x = 21
+    Cfg.commands.num_bins_vel_y = 1
+    Cfg.commands.num_bins_vel_yaw = 21
+    Cfg.commands.num_bins_body_height = 1
+    Cfg.commands.num_bins_gait_frequency = 1
+    Cfg.commands.num_bins_gait_phase = 1
+    Cfg.commands.num_bins_gait_offset = 1
+    Cfg.commands.num_bins_gait_bound = 1
+    Cfg.commands.num_bins_gait_duration = 1
+    Cfg.commands.num_bins_footswing_height = 1
+    Cfg.commands.num_bins_body_roll = 1
+    Cfg.commands.num_bins_body_pitch = 1
+    Cfg.commands.num_bins_stance_width = 1
+
+    Cfg.normalization.friction_range = [0, 1]
+    Cfg.normalization.ground_friction_range = [0, 1]
+    Cfg.terrain.yaw_init_range = 3.14
+    Cfg.normalization.clip_actions = 10.0
+
+    Cfg.commands.exclusive_phase_offset = False
+    Cfg.commands.pacing_offset = False
+    Cfg.commands.binary_phases = True
+    Cfg.commands.gaitwise_curricula = True
+
+    env = VelocityTrackingEasyEnv(sim_device='cuda:0', headless=False, cfg=Cfg)
+
+    # log the experiment parameters
+    logger.log_params(AC_Args=vars(AC_Args), PPO_Args=vars(PPO_Args), RunnerArgs=vars(RunnerArgs),
+                      Cfg=vars(Cfg))
+
+    env = HistoryWrapper(env)
+    gpu_id = 0
+    runner = Runner(env, device=f"cuda:{gpu_id}")
+    runner.learn(num_learning_iterations=100000, init_at_random_ep_len=True, eval_freq=100)
+
+
+if __name__ == '__main__':
+    from pathlib import Path
+    from ml_logger import logger
+    from go1_gym import MINI_GYM_ROOT_DIR
+
+    stem = Path(__file__).stem
+    logger.configure(logger.utcnow(f'gait-conditioned-agility/%Y-%m-%d/{stem}/%H%M%S.%f'),
+                     root=Path(f"{MINI_GYM_ROOT_DIR}/runs").resolve(), )
+    logger.log_text("""
+                charts: 
+                - yKey: train/episode/rew_total/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_tracking_lin_vel/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_tracking_contacts_shaped_force/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_action_smoothness_1/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_action_smoothness_2/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_tracking_contacts_shaped_vel/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_orientation_control/mean
+                  xKey: iterations
+                - yKey: train/episode/rew_dof_pos/mean
+                  xKey: iterations
+                - yKey: train/episode/command_area_trot/mean
+                  xKey: iterations
+                - yKey: train/episode/max_terrain_height/mean
+                  xKey: iterations
+                - type: video
+                  glob: "videos/*.mp4"
+                - yKey: adaptation_loss/mean
+                  xKey: iterations
+                """, filename=".charts.yml", dedent=True)
+
+    # to see the environment rendering, set headless=False
+    train_go1(headless=False)