OARL/main.py at master · TMIS-Turbo/OARL · GitHub

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
import numpy as np
import pandas as pd
import random as rn
import argparse
import os
import gym
import matplotlib.pyplot as plt
import Environment.environment
import oarl
import torch


# Configurations
parser = argparse.ArgumentParser(description='RL algorithms with PyTorch in Pendulum environment')
parser.add_argument('--env', type=str, default='highway-v0', help='CartPole environment')
parser.add_argument('--algo', type=str, default='oarl', help='select an algorithm among sac, asac')
parser.add_argument('--seed', type=int, default=0, help='seed for random number generators')
parser.add_argument('--episodes', type=int, default=400, help='training episode number')
parser.add_argument('--max_step', type=int, default=200, help='max episode step')
parser.add_argument('--state_dim', type=int, default=16, help='state dimension')
parser.add_argument('--action_dim', type=int, default=1, help='action dimension')
parser.add_argument('--action_numb', type=int, default=3, help='action number')
parser.add_argument('--mode', type=str, default='train', help='train')
parser.add_argument('--save_dir_model', type=str, default='model/', help='the path to save models')
parser.add_argument('--save_dir_data', type=str, default='result/', help='the path to save data')
parser.add_argument('--save_dir_train_data', type=str, default='train/', help='the path to save training data')
args = parser.parse_args()

# Set environment
env = gym.make(args.env)

# Set a random seed
env.seed(args.seed)
np.random.seed(args.seed)
rn.seed(args.seed)
torch.manual_seed(args.seed)


def train():
    env.start(gui=False)
    model_l = oarl.Agent(args.state_dim, args.action_dim, args.action_numb)

    if not os.path.exists(args.save_dir_model):
        os.mkdir(args.save_dir_model)
    model_l.train()
    print("The model is training")

    score = 0.0
    total_reward = []
    episode = []
    print_interval = 10
    train_interval = 2
    interaction_times = 0

    v = []
    v_epi = []
    v_epi_mean = []

    ax = []
    ax_epi = []
    ax_epi_mean = []

    ay = []
    ay_epi = []
    ay_epi_mean = []

    cn = 0.0
    cn_epi = []

    for n_epi in range(args.episodes):
        s = env.reset()
        done = False
        step_number = 0
        while not (done or step_number == args.max_step):
            s = np.array(s, dtype=float)

            a_l = model_l.select_action_single(s, args.mode)
            s_prime, r, done, _, _, _, _ = env.step(a_l)

            model_l.replay_buffer.add(s, a_l, r, s_prime, done)

            s = s_prime
            step_number += 1
            interaction_times += 1
            score += r

            if args.mode == "train" and n_epi > print_interval and interaction_times % train_interval == 0:
                model_l.train_model()

            v.append(s[0]*35)
            v_epi.append(s[0]*35)
            ax.append(s[-2]*10)
            ax_epi.append(s[-2]*10)
            ay_ = (s[0]*35)*(s[-1]*10*3.14/180)
            ay.append(ay_)
            ay_epi.append(ay_)
            d_f = 100 * s[2]
            d_b = 100 * s[4]
            if d_f < 3 or d_b < 2.5:
                cn += 1

        if n_epi % print_interval == 0 and n_epi != 0:
            print("episode :{}, avg score_v : {:.1f}, interaction_times:{}".format(n_epi, score/print_interval, interaction_times))

            episode.append(n_epi)
            total_reward.append(score / print_interval)
            cn_epi.append(cn)

            v_mean = np.mean(v_epi)
            v_epi_mean.append(v_mean)

            ax_mean = np.mean(ax_epi)
            ax_epi_mean.append(ax_mean)

            ay_mean = np.mean(ay_epi)
            ay_epi_mean.append(ay_mean)

            score = 0.0

            v_epi = []
            ax_epi = []
            ay_epi = []
            cn = 0.0

        if args.mode == "train" and (n_epi+1) % 100 == 0:
            model_l.save_model(n_epi+1, args.save_dir_model)

    df = pd.DataFrame([])
    df["n_epi"] = episode
    df["total_reward"] = total_reward
    df["v_epi_mean"] = v_epi_mean
    df["ax_epi_mean"] = ax_epi_mean
    df["ay_epi_mean"] = ay_epi_mean
    df["cn_epi"] = cn_epi

    df_ = pd.DataFrame([])
    df_["v"] = v
    df_["ax"] = ax
    df_["ay"] = ay

    if not os.path.exists(args.save_dir_data):
        os.mkdir(args.save_dir_data)
    train_data_path = os.path.join(args.save_dir_data, args.save_dir_train_data)
    if not os.path.exists(train_data_path):
        os.mkdir(train_data_path)

    df.to_csv('./' + train_data_path + '/train_rac.csv', index=0)
    df_.to_csv('./' + train_data_path + '/train_rac_.csv', index=0)

    plt.plot(episode, total_reward)
    plt.xlabel('episode')
    plt.ylabel('total_reward')
    plt.show()

    env.close()


if __name__ == "__main__":
    train()