swangnice/rotor_py_control
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Switched to cmd_motor_speeds abstraction.

Browse Source
This commit is contained in:
spencerfolk
2024-01-04 12:17:41 -05:00
parent 69682680ef
commit 6856a4f146
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
examples/ppo_hover_train.py
View File

@@ -47,7 +47,7 @@ reward_function = lambda obs, act: hover_reward(obs, act, weights={'x': 1, 'v':
# Make the environment. For this demo we'll train a policy to command collective thrust and body rates. # Make the environment. For this demo we'll train a policy to command collective thrust and body rates.
# Turning render_mode="None" will make the training run much faster, as visualization is a current bottleneck. # Turning render_mode="None" will make the training run much faster, as visualization is a current bottleneck.
env = gym.make("Quadrotor-v0", env = gym.make("Quadrotor-v0",
control_mode ='cmd_ctbr', control_mode ='cmd_motor_speeds',
reward_fn = reward_function, reward_fn = reward_function,
quad_params = quad_params, quad_params = quad_params,
max_time = 5, max_time = 5,
@@ -74,7 +74,7 @@ epoch_count = 0
while True: # Run indefinitely.. while True: # Run indefinitely..
# This line will run num_timesteps for training and log the results every so often. # This line will run num_timesteps for training and log the results every so often.
model.learn(total_timesteps=num_timesteps, reset_num_timesteps=False, tb_log_name="PPO-Quad_"+start_time.strftime('%H-%M-%S')) model.learn(total_timesteps=num_timesteps, reset_num_timesteps=False, tb_log_name="PPO-Quad_cmd-motor_"+start_time.strftime('%H-%M-%S'))
# Save the model # Save the model
model.save(f"{models_dir}/PPO/{start_time.strftime('%H-%M-%S')}/hover_{num_timesteps*(epoch_count+1)}") model.save(f"{models_dir}/PPO/{start_time.strftime('%H-%M-%S')}/hover_{num_timesteps*(epoch_count+1)}")