forked from IgnacioCarlucho/deepPID
-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.py
181 lines (148 loc) · 7.48 KB
/
main.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
from replay_buffer import ReplayBuffer
import tensorflow as tf
import numpy as np
from ou_noise import OUNoise
from iddpg import IDDPG
from td3 import TD3
from robots import pioneer_pi
# if in windows
# import matplotlib
# matplotlib.use('Agg')
import matplotlib.pyplot as plt
from utils import str2bool
import argparse
ACTOR_LEARNING_RATE = 0.0001
CRITIC_LEARNING_RATE = 0.0001
# Soft target update param
TAU = 0.001
parser = argparse.ArgumentParser('deepid')
parser.add_argument('--gpu', type=str, choices=['gpu', 'cpu'], default='cpu')
parser.add_argument('--alg', type=str, choices=['iddpg', 'ddpg', 'itd3'], default='iddpg')
parser.add_argument('--epochs', type=int, default=500)
parser.add_argument('--sim', type=int, default=100)
parser.add_argument('--epsilon', type=float, default=1.)
parser.add_argument("--train", type=str2bool, nargs='?', const=True, default=True, help="Activate train mode.")
parser.add_argument("--img", type=str2bool, nargs='?', const=True, default=True, help="Activate save image.")
parser.add_argument("--reset", type=str2bool, nargs='?', const=True, default=True, help="resets after each episode.")
parser.add_argument("--simulation", type=str2bool, nargs='?', const=True, default=True, help="Is this run on gazebo or in the real vehicel.")
parser.add_argument("--load", type=str2bool, nargs='?', const=True, default=False, help="Loads policy.")
parser.add_argument('--epsilon_decay', type=float, default=0.0002)
parser.add_argument('--psi', type=float, default=1.)
parser.add_argument('--pid', type=str, choices=['pid', 'pi'], default='pi')
parser.add_argument('--max_action', type=float, default=50.)
parser.add_argument('--min_action', type=float, default=0.0001)
parser.add_argument('--seed', type=int, default=51234)
parser.add_argument('--save_mod', type=int, default=500)
args = parser.parse_args()
alg = args.alg
DEVICE = args.gpu
max_action = args.max_action
min_action = args.min_action
epochs = args.epochs
epsilon = args.epsilon
min_epsilon = 0.1
decay_rate = args.epsilon_decay # 0.9/ num of episodes until min_epsilon (0.1)
BUFFER_SIZE = 100000
RANDOM_SEED = args.seed
MINIBATCH_SIZE = 64
PSI = args.psi
save_mod = args.save_mod
with tf.Session() as sess:
np.random.seed(RANDOM_SEED)
tf.set_random_seed(RANDOM_SEED)
state_dim = 20 + 2 +4
action_dim = 4
robot = pioneer_pi("pioneer", n_actions=action_dim,save_image=False, dt=0.1, Teval = 1, simulation=args.simulation,reset=args.reset, ep_length = args.sim)
if alg == 'iddpg':
low = IDDPG(sess, state_dim, action_dim, max_action, min_action, ACTOR_LEARNING_RATE,CRITIC_LEARNING_RATE, TAU, RANDOM_SEED, device=DEVICE)
elif alg == 'itd3':
low = TD3(sess, state_dim, action_dim, max_action, min_action, ACTOR_LEARNING_RATE,CRITIC_LEARNING_RATE, TAU, RANDOM_SEED, device=DEVICE)
else:
raise NotImplementedError
sess.run(tf.global_variables_initializer())
if args.load:
low.load()
replay_buffer = ReplayBuffer(BUFFER_SIZE, RANDOM_SEED)
ruido = OUNoise(action_dim, mu = 0.0)
total_ep_reward = np.zeros(epochs)
for i in range(epochs):
# define goal
# option 1: sample from especific velocities
sample_vx = np.array([0.,.1,.2,.3,.4,.5,.15,.25,.35,.45,0.,-.1,-.2,-.3,-.4,-.5,-.15,-.25,-.35,-.45])
sample_wz = np.array([0.,.1,.2,.3,.15,.25,0.,-.1,-.2,-.3,-.15,-.25,])
velocity_req = np.array([np.random.choice(sample_vx), np.random.choice(sample_wz)])
# option 2: all random (problem is you get velocities like 0.08 which make no sense)
# velocity_req = np.array([np.random.rand(1)[0] -.5, 0.4*np.random.rand(1)[0] -.2])
# Option 3: Define a specific velocity
# velocity_req = np.array([0.21,0.11])
# Initial state
action = np.zeros(action_dim)
position, velocities, u = robot.reset() # velocities = np.zeros((10,6))
robot.get_set_point(velocity_req)
velocity_error = np.subtract(velocities[9], velocity_req)
state = np.reshape(np.vstack((velocity_error,velocities,np.reshape(action,(2,2)))),(state_dim,))
done = False
epsilon -= decay_rate
epsilon = np.maximum(min_epsilon,epsilon)
episode_r = 0.
step = 0
action_buffer = np.zeros(((args.sim+1),action_dim))
velocity_buffer = np.zeros(((args.sim+1)*10,2))
r_vector = np.zeros(args.sim+1)
while (not done):
action = low.predict_action(np.reshape(state,(1,state_dim)))[0]
action = np.clip(action,min_action,max_action)
action = action + max(4.*epsilon,0)*ruido.noise()
action = np.clip(action,min_action,max_action)
new_position, new_velocities, u = robot.run(action)
# this should all go to the robot.run
velocity_error = np.subtract(new_velocities[9], velocity_req)
next_state = np.reshape(np.vstack((new_velocities,velocity_error,np.reshape(action,(2,2)))),(state_dim,))
reward, done = robot.get_reward_v2(new_velocities,velocity_req,step)
action_buffer[step] = action
j = step*10
velocity_buffer[j] = new_velocities[0]
velocity_buffer[j+1] = new_velocities[1]
velocity_buffer[j+2] = new_velocities[2]
velocity_buffer[j+3] = new_velocities[3]
velocity_buffer[j+4] = new_velocities[4]
velocity_buffer[j+5] = new_velocities[5]
velocity_buffer[j+6] = new_velocities[6]
velocity_buffer[j+7] = new_velocities[7]
velocity_buffer[j+8] = new_velocities[8]
velocity_buffer[j+9] = new_velocities[9]
r_vector[step] = reward
replay_buffer.add(np.reshape(state, (state_dim,)), np.reshape(action, (action_dim,)), reward,
done, np.reshape(next_state, (state_dim,)))
state = next_state
step += 1
episode_r = episode_r + reward
if replay_buffer.size() > MINIBATCH_SIZE:
s_batch, a_batch, r_batch, t_batch, s2_batch = replay_buffer.sample_batch(MINIBATCH_SIZE)
low.train(s_batch, a_batch, r_batch, t_batch, s2_batch,MINIBATCH_SIZE)
print(i, step, 'last r', round(reward,3), 'epsilon', round(epsilon,3),'episode reward','**',round(episode_r,3),'**' )
print('req', velocity_req, 'last v', np.round(new_velocities[9],3))
ruido.reset()
robot.reset()
total_ep_reward[i] = episode_r
# np.save('velocity-buff.np',velocity_buffer)
if i%save_mod==0:
np.save('figs/velocity_buffer' + str(i), velocity_buffer)
np.save('figs/action_buffer' + str(i), action_buffer)
plt.plot(action_buffer)
plt.legend(['kp1','ki1','kp2','ki2'])
plt.savefig('figs/action' + str(i) + '.png')
# plt.savefig('action.png')
# plt.show()
plt.clf()
plt.plot(velocity_buffer)
plt.legend(['vx = ' + str(np.round(robot.set_point[0],2)) ,'wz = '+ str(np.round(robot.set_point[1],2))])
plt.savefig('figs/velocity' + str(i) + '.png')
# plt.savefig('velocity.png')
# plt.show()
plt.clf()
np.save('velocity-buff',velocity_buffer)
np.save('action_buf',action_buffer)
np.save('r_vector',r_vector)
print('total_av_reward' , np.mean(total_ep_reward))
low.save()