forked from ZikangXiong/VRL_CodeReview
-
Notifications
You must be signed in to change notification settings - Fork 0
/
tape.py
73 lines (61 loc) · 2.74 KB
/
tape.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
from main import *
from Environment import Environment
from DDPG import *
from shield import Shield
import argparse
def tape (learning_method, number_of_rollouts, simulation_steps,learning_eposides, critic_structure, actor_structure, train_dir,\
nn_test=False, retrain_shield=False, shield_test=False, test_episodes=100):
A = np.matrix([[5.5197e-17,-3.5503e-17,6.2468e-32],
[2.7756e-17,0,0],
[0,2.7756e-17,0]
])
B = np.matrix([[0.25],
[0],
[0]
])
#intial state space
s_min = np.array([[-1.0],[-1.0], [-1.0]])
s_max = np.array([[ 1.0],[ 1.0], [ 1.0]])
Q = np.matrix("1 0 0 ; 0 1 0; 0 0 1")
R = np.matrix(".0005")
x_min = np.array([[-3],[-3],[-3]])
x_max = np.array([[ 3],[ 3], [3]])
u_min = np.array([[-10.]])
u_max = np.array([[ 10.]])
env = Environment(A, B, u_min, u_max, s_min, s_max, x_min, x_max, Q, R)
args = { 'actor_lr': 0.0001,
'critic_lr': 0.001,
'actor_structure': actor_structure,
'critic_structure': critic_structure,
'buffer_size': 1000000,
'gamma': 0.99,
'max_episode_len': 1,
'max_episodes': learning_eposides,
'minibatch_size': 64,
'random_seed': 6553,
'tau': 0.005,
'model_path': train_dir+"model.chkp",
'enable_test': nn_test,
'test_episodes': test_episodes,
'test_episodes_len': 500}
actor = DDPG(env, args)
#################### Shield #################
model_path = os.path.split(args['model_path'])[0]+'/'
linear_func_model_name = 'K.model'
model_path = model_path+linear_func_model_name+'.npy'
shield = Shield(env, actor, model_path, force_learning=retrain_shield, debug=False)
shield.train_shield(learning_method, number_of_rollouts, simulation_steps, eq_err=0, explore_mag = 0.5, step_size = 0.5)
if shield_test:
shield.test_shield(test_episodes, 500, mode="single")
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='Running Options')
parser.add_argument('--nn_test', action="store_true", dest="nn_test")
parser.add_argument('--retrain_shield', action="store_true", dest="retrain_shield")
parser.add_argument('--shield_test', action="store_true", dest="shield_test")
parser.add_argument('--test_episodes', action="store", dest="test_episodes", type=int)
parser_res = parser.parse_args()
nn_test = parser_res.nn_test
retrain_shield = parser_res.retrain_shield
shield_test = parser_res.shield_test
test_episodes = parser_res.test_episodes if parser_res.test_episodes is not None else 100
tape("random_search", 100, 50, 0, [240,200], [280,240,200], "ddpg_chkp/tape/240200280240200/", nn_test=nn_test, retrain_shield=retrain_shield, shield_test=shield_test, test_episodes=test_episodes)