-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfixed_env_wrap_fft.py
196 lines (165 loc) · 6.87 KB
/
fixed_env_wrap_fft.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
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
import LiveStreamingEnv.fixed_env as fixed_env
import LiveStreamingEnv.load_trace as load_trace
import numpy as np
from env_args import EnvArgs
class FixedEnvWrap(fixed_env.Environment):
def __init__(self, video_file_id, trace='test'):
self.args = EnvArgs()
all_cooked_time, all_cooked_bw, all_file_names = load_trace.load_trace(self.args.bw_trace[trace])
super().__init__(all_cooked_time=all_cooked_time,
all_cooked_bw=all_cooked_bw,
random_seed=self.args.random_seed,
VIDEO_SIZE_FILE =self.args.video_size_files[video_file_id],
logfile_path='./log/',
Debug=False)
# self.state = np.zeros((args.s_info, args.s_len)) # state info for past frames
self.state_gop = np.zeros((self.args.s_gop_info, self.args.s_gop_len)) # state info for past gops
self.last_bit_rate = 0
self.reward_gop = 0
self.last_reward_gop = 0
self.action_map = self._set_action_map()
# state info for gop
self.gop_time_interval = 0
self.gop_size = 0
self.next_gop_sizes = 0
self.gop_delay = 0
# record finer level thps, thp/0.5s
self.frame_thps = [0] * 16
self.last_frame_thp = 0.0
# info for traces
self.traces_len = len(all_file_names)
@staticmethod
def _set_action_map():
bit_rate_levels = [0, 1, 2, 3]
target_buffer_levels = [0, 1]
action_map = []
for bitrate_idx in range(len(bit_rate_levels)):
for target_buffer_idx in range(len(target_buffer_levels)):
action_map.append((bit_rate_levels[bitrate_idx], target_buffer_levels[target_buffer_idx]))
return action_map
# @staticmethod
def _predict_gop_sizes(self):
# predict next gop sizes based on past gop sizes
if self.last_bit_rate == 0:
return self.gop_size * np.array([1, 1.7, 1.7*1.4, 1.7*1.4*1.5])
elif self.last_bit_rate == 1:
return self.gop_size * np.array([1/1.7, 1, 1.4, 1.4*1.5])
elif self.last_bit_rate == 2:
return self.gop_size * np.array([1/(1.7*1.4), 1/1.4, 1, 1.5])
else:
return self.gop_size * np.array([1/(1.7*1.4*1.5), 1/(1.4*1.5), 1/1.5, 1])
# @staticmethod
def _get_next_gop_sizes(self, cdn_has_frame):
gop_flags = cdn_has_frame[4]
# print(gop_flags)
# if there are no frames on cdn, predict future gop sizes
len_cdn = len(cdn_has_frame[4])
if len_cdn == 0:
# print('cdn has no frames')
return self._predict_gop_sizes()
# if there are more than one gop on cdn
# assert gop_flags[0] == 1
total_gop_flag = False
gop_start_id = 0
gop_end_id = 0
for i in range(0, len(gop_flags)):
if gop_flags[i] == 1:
gop_start_id = i
break
for i in range(gop_start_id+1, len(gop_flags)):
if gop_flags[i] == 1:
total_gop_flag = True
gop_end_id = i
break
# print("gop start id is : %d" % (gop_start_id))
# if total_gop_flag == False:
# gop_end_id = len(gop_flags)
# print("gop end id is : %d" % (gop_end_id))
if total_gop_flag == True:
next_frame_sizes = np.array(cdn_has_frame)[:4, gop_start_id:gop_end_id]
next_gop_sizes = np.sum(next_frame_sizes, axis=1)
# print('there is a whole gop on cdn')
return next_gop_sizes
else:
next_I_frame_sizes = np.array(cdn_has_frame)[:4, gop_start_id:gop_start_id+1].reshape(4)
next_P_frame_average = np.mean(np.array(cdn_has_frame)[:4, gop_start_id:-1], axis=1)
# print(next_I_frame_sizes)
# print(next_P_frame_average)
if len(cdn_has_frame[4]) == 1:
next_P_frame_average = np.array([5000, 5000*1.7, 5000*1.7*1.4, 5000*1.7*1.4*1.5])
next_gop_sizes = next_I_frame_sizes + next_P_frame_average * 49
# print('there is part of gop on cdn')
# print(next_I_frame_sizes)
# print(next_P_frame_average)
# print(next_gop_sizes)
return next_gop_sizes
# return gop state
def step_gop(self, action):
bit_rate, target_buffer = self.action_map[action]
time, time_interval, send_data_size, frame_len, rebuf,\
buffer_size, play_time_len,end_delay, cdn_newest_id,\
download_id, cdn_has_frame, decision_flag, buffer_flag,cdn_flag, end_of_video =\
self.get_video_frame(bit_rate, target_buffer)
# reward setting
if not cdn_flag:
reward_frame = self.args.frame_time_len * float(self.args.bitrate[bit_rate]) / 1000 \
- self.args.rebuf_penalty * rebuf \
- self.args.latency_penalty * end_delay
else:
reward_frame = -(self.args.rebuf_penalty * rebuf)
# collect finer level thps / interval > 0.5s
if not cdn_flag and time_interval > 0:
frame_thp = send_data_size / time_interval
if frame_thp != self.last_frame_thp:
self.frame_thps.pop(0)
self.frame_thps.append(frame_thp / 1000000)
self.last_frame_thp = frame_thp
if not decision_flag:
self.reward_gop += reward_frame
# collect frames info in last gop
if not cdn_flag: # if cdn is rebuffering, then time interval is not download time
self.gop_time_interval += time_interval
self.gop_size += send_data_size # last gop size
self.gop_delay += end_delay # last gop size
if decision_flag or end_of_video:
reward_frame += -1 * self.args.smooth_penalty * (abs(self.args.bitrate[bit_rate] - self.args.bitrate[self.last_bit_rate]) / 1000)
self.last_bit_rate = bit_rate
self.reward_gop += reward_frame # the last frame in a gop
self.last_reward_gop = self.reward_gop
self.reward_gop = 0 # reset reward gop as 0
# calculate next gop sizes for 4 bitrate levels [500k, 800k, 1200k, 1800k]
self.next_gop_sizes = self._get_next_gop_sizes(cdn_has_frame)
# collect gop state info
self.state_gop = np.roll(self.state_gop, -1, axis=1)
self.state_gop[0, -1] = buffer_size # current buffer size [0, 10] [fc]
self.state_gop[1, -1] = self.args.bitrate[bit_rate] / 1000 # last bitrate [0, 2] [fc]
# self.state_gop[2, -1] = self.gop_size / 1000000 / max(self.gop_time_interval, 1e-6) # last throughput Mbps [0, 10] [conv]
self.state_gop[2, :] = self.frame_thps # last throughput Mbps [0, 10] [conv]
self.state_gop[3, -1] = self.gop_delay / 100 # gop delay (100ms) [conv]
self.state_gop[4, -1] = (1 if buffer_flag else 0) # if True, no buffering content, should choose target buffer as 0. [fc]
self.state_gop[5, -1] = (1 if cdn_flag else 0) # if True, cdn has no content. [fc]
self.state_gop[6, :4] = self.next_gop_sizes / 1000000 # gop size (Mb) [0, 10] [conv]
# self.state_gop[7, :] = self.frame_thps # finer level thps
# test new features, FFT of thps
_fft = np.fft.fft(self.frame_thps)
self.state_gop[7, :] = _fft.real # [conv]
self.state_gop[8, :] = _fft.real # [conv]
# reset gop info
self.gop_time_interval = time_interval
self.gop_size = send_data_size
self.gop_delay = end_delay
return reward_frame, end_of_video, decision_flag
def get_reward_gop(self):
return self.last_reward_gop
def get_state_gop(self):
return self.state_gop
def reset(self):
self.state_gop = np.zeros((self.args.s_gop_info, self.args.s_gop_len))
self.last_bit_rate = 0
self.reward_gop = 0
self.last_reward_gop = 0
self.gop_time_interval = 0
self.gop_size = 0
self.next_gop_sizes = 0
self.gop_delay = 0
return self.state_gop