def train():
runtime = 5. # time limit of the episode
init_pose = np.array([0., 0., 4.0, 0., 0., 0.0]) # initial pose
init_velocities = np.array([0., 0., 0.0]) # initial velocities
init_angle_velocities = np.array([0., 0., 0.]) # initial angle velocities
file_output = 'rewards.txt' # file name for saved results
num_episodes = 10
target_pos = np.array([0., 0., 40.])
task = Task(init_pose=init_pose,
init_velocities=init_velocities,
init_angle_velocities=init_angle_velocities,
target_pos=target_pos)
agent = DDPG(task)
labels = ['episod', 'avg_reward', 'total_reward']
results = {x: [] for x in labels}
with open(file_output, 'w') as csvfile:
writer = csv.writer(csvfile)
writer.writerow(labels)
best_total_reward = -1000
for i_episode in range(1, num_episodes + 1):
state = agent.reset_episode() # start a new episode
total_reward = 0
rewards = []
while True:
# select action according to the learned policy and the exploration noise
action = agent.act(state)
# execute the action and observe the reward and the next state
next_state, reward, done = task.step(action)
# sample mini batch and learn
agent.step(action, reward, next_state, done)
# data tracking
total_reward += reward
rewards.append(reward)
if total_reward > best_total_reward:
best_total_reward = total_reward
state = next_state
if done:
avg_reward = np.mean(np.array(rewards))
print(task.sim.pose)
#to_write = [task.sim.time] + list(task.sim.pose) + list(task.sim.v) + list(task.sim.angular_v) + list(rotor_speeds)
#for ii in range(len(labels)):
# results[labels[ii]].append(to_write[ii])
#writer.writerow(to_write)
to_write = [i_episode] + [avg_reward] + [total_reward]
for ii in range(len(labels)):
results[labels[ii]].append(to_write[ii])
print(
"\rEpisode = {:4d}, total_reward = {:7.3f}, avg_reward={:7.3} (best = {:7.3f})"
.format(i_episode, total_reward, avg_reward,
best_total_reward),
end="") # [debug]
break
sys.stdout.flush()
return agent
num_episodes = 1000
init_pose = np.array([0., 0., 0., 0., 0., 0.])
target_pos = np.array([0., 0., 10.])
init_velocities = np.array([0., 0., 0.]) # initial velocities
init_angle_velocities = np.array([0., 0., 0.])
task = Task(init_pose=init_pose,
target_pos=target_pos,
init_angle_velocities=init_angle_velocities,
init_velocities=init_velocities)
best_score = -np.inf
agent = DDPG(task)
for i_episode in range(1, num_episodes + 1):
state = agent.reset_episode() # start a new
score = 0
while True:
action = agent.act(state)
next_state, reward, done = task.step(action)
agent.step(action, reward, next_state, done)
state = next_state
score += reward
best_score = max(best_score, score)
if done:
print("\rEpisode = {:4d}, score = {:7.3f} (best = {:7.3f})".format(
i_episode, score, best_score),
end="") # [debug]
break
sys.stdout.flush()