Methods and apparatuses for feature-driven machine-to-machine communications
地域：
Ottawa
摘要
Methods and apparatuses for feature-driven machine-to-machine communications are described. At a feature encoder, features are extracted from sensed raw information, to generate features that compress the raw information by a compression ratio. The feature encoder implements a probabilistic encoder to generate the features, each feature providing information about a respective probability distribution that each represents one or more aspects of the subject. The probabilistic encoder is designed to provide a compression ratio that satisfies a predetermined physical channel capacity limit for a transmission channel. The features are transmitted over the transmission channel.
說明書
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
The present disclosure relates to machine-to-machine (M2M) communications in a network. In particular, the present disclosure relates to apparatuses and methods for feature-driven compression for M2M communications.
Machine-to-machine (M2M) communications is a technology that realizes a network for collecting information from devices (e.g., sensors, smart meters, Internet of Things (IoT) devices, and/or other low-end devices) that are typically massively and densely deployed, and for transmitting information captured by those devices to other applications in the network. M2M networks may be wired or wireless and may have a relatively large geographical distribution (e.g., across a country or across the world). M2M communications typically do not involve direct human intervention for information collection.
Developments for 5G New Radio (NR) system include features to support massive machine type communications (mMTC) that connects large numbers (e.g., millions or billions) of IoT equipment by a wireless system. It is expected in the near future that the amount of M2M communications conducted over-the-air will bypass those of human-related communications. Accordingly, it would be desirable to provide ways to better support large-scale M2M communications.
權利要求
The invention claimed is:1. An apparatus for feature-based compression, the apparatus comprising:a feature encoder configured to extract one or more features from a set of raw information, to generate a set of features that compresses the raw information by a compression ratio that satisfies a predetermined physical channel capacity limit for a transmission channel, each feature providing information about a respective probability distribution that each represents one or more aspects of the subject; and a transmitter configured to transmit a reduced set of the features over the transmission channel. 2. The apparatus of claim 1 , wherein the feature encoder implements a probabilistic encoder to generate the set of features and the probabilistic encoder is implemented using an encoder deep neural network (DNN), and wherein the encoder DNN is trained to satisfy: a first target of maximizing likelihood between a set of recovered information at a corresponding decoder DNN, and a second target of minimizing an upper boundary of mutual information to be within the predetermined physical channel capacity limit. 3. The apparatus of claim 2 , wherein the encoder DNN and the decoder DNN are trained together. 4. The apparatus of claim 2 , wherein the compression ratio provided by the trained encoder DNN and the decoder DNN has been determined by performing training on a plurality of candidate encoder and decoder DNN pairs, each candidate encoder and decoder DNN pair providing a respective different compression ratio, and selecting the candidate encoder and decoder DNN pair and associated compression ratio that minimizes the upper boundary of mutual information. 5. The apparatus of claim 1 , wherein the reduced set of features omit a correlated feature of the set of features. 6. The apparatus of claim 5 , further comprising:a historical database storing at least one previously transmitted feature; and wherein the correlated feature is any feature that is unchanged compared to the at least one previously transmitted feature. 7. The apparatus of claim 5 , wherein the correlated feature is indicated by a control message. 8. The apparatus of claim 1 , wherein the transmitter is configured to:assign a sub-channel for transmission of each respective feature, the assigning being based on a relative importance of each feature; and transmit the set of features over the sub-channels. 9. The apparatus of claim 8 wherein each feature indicates an expectation value of the respective probability distribution and a variance value of the respective probability distribution, and the relative importance of each feature is determined based on the variance value of each respective feature. 10. The apparatus of claim 9 wherein the transmitter is further configured to:select a transmission scheme for each assigned sub-channel, the transmission scheme being selected to indicate the variance value of the feature assigned to each respective sub-channel; and transmit the expectation value of each feature over the respective sub-channel in accordance with the respective transmission scheme. 11. The apparatus of claim 10 wherein the transmitter is further configured to:generate a control message or header indicating the selected transmission scheme and assigned sub-channel for each feature; and transmit the control message or header. 12. The apparatus of claim 9 , wherein a first feature having a first variance value and a second feature having a second variance value similar to the first variance value are assigned to the same sub-channel for transmission. 13. A method for managing a plurality of sensors monitoring a common subject, each sensor generating and transmitting a respective set of features representing one or more aspects of the subject, the method comprising:determining a correlated feature that is highly correlated between a first set of features generated by a first sensor and a second set of features generated by a second sensor; generating a control message to the first sensor to cause the first sensor to omit the correlated feature from transmission; and reconstructing the first set of features from a transmission from the first sensor by filling in the omitted correlated feature. 14. The method of claim 13 , wherein the first set of features is reconstructed by copying the correlated feature from the second set of features received from the second sensor. 15. The method of claim 13 , further comprising:determining that the correlated feature is a background feature that is unchanged over a predetermined time period; and wherein the first set of features is reconstructed by copying the correlated feature from a historical database containing a previously transmitted instance of the background feature. 16. The method of claim 15 , wherein a same or different control message is generated to cause the first sensor and the second sensor to omit the background feature from transmission. 17. The method of claim 13 , further comprising:determining a requested set of features that is requested by an application, wherein the requested set of features is a subset of the first set of features; generating the control message to the first sensor to cause the first sensor to transmit only the subset of features; and reconstructing the first set of features from a transmission from the first sensor by filling in untransmitted features with random values. 18. The method of claim 13 , wherein a same or different control message is generated to cause the first sensor and the second sensor to alternately transmit or omit the correlated feature. 19. The method of claim 13 , wherein all features in the first set of features are highly correlated with the second set of features, and wherein the control message causes the first sensor to enter a sleep mode. 20. The method of claim 13 , wherein the first sensor has a poorer physical layer transmission performance than the second sensor.
意見反饋