< draft-choi-icnrg-aiot-00.txt   draft-choi-icnrg-aiot-01.txt >
ICN Research Group J.K.Choi ICN Research Group J.K.Choi
Internet-Draft N.K.Kim Internet-Draft N.K.Kim
Intended status: Informational J.S.Han Intended status: Informational J.S.Han
Expires: September 12, 2019 M.K.Kim Expires: January 8, 2020 M.K.Kim
KAIST KAIST
G.M.Lee July 7, 2019
Liverpool John Moores University
March 11, 2019
Requirements and Challenges for User-level Service Managements of IoT
Network by utilizing Artificial Intelligence
draft-choi-icnrg-aiot-00 Requirements and Challenges for User-level Service Managements of IoT
Network by utilizing Artificial Intelligence
draft-choi-icnrg-aiot-01
Abstract Abstract
This document describes the requirements and challenges to employ This document describes the requirements and challenges to employ artificial intelligence
artificial intelligence (AI) into the constraint Internet of Things (AI) into the constraint Internet of Things (IoT) service environment for embedding
(IoT) service environment for embedding intelligence and increasing intelligence and increasing efficiency.
efficiency. The IoT service environment includes heterogeneous and multiple IoT devices and systems
that work together in a cooperative and intelligent way to manage homes, buildings, and
The IoT service environment includes heterogeneous and multiple IoT complex autonomous systems. Therefore, it is becoming very essential to integrate IoT and
devices and systems that work together in a cooperative and AI technologies to increase the synergy between them. However, there are several
intelligent way to manage homes, buildings, and complex autonomous limitations to achieve AI enabled IoT as the availability of IoT devices is not always high,
systems. Therefore, it is becoming very essential to integrate IoT and IoT networks cannot guarantee a certain level of performance in real-time applications
and AI technologies to increase the synergy between them. However, due to resource constraints.
there are several limitations to achieve AI enabled IoT as the This document intends to present a right direction to empower AI in IoT for learning and
availability of IoT devices is not always high, and IoT networks analyzing the usage behaviors of IoT devices/systems and human behaviors based on
cannot guarantee a certain level of performance in real-time previous records and experiences. With AI enabled IoT, the IoT service environment can be
applications due to resource constraints. intelligently managed in order to compensate for the unexpected performance degradation
often caused by abnormal situations.
This document intends to present a right direction to empower AI in
IoT for learning and analyzing the usage behaviors of IoT
devices/systems and human behaviors based on previous records and
experiences. With AI enabled IoT, the IoT service environment can be
intelligently managed in order to compensate for the unexpected
performance degradation often caused by abnormal situations.
Status of This Memo This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
This Internet-Draft is submitted in full conformance with the Internet-Drafts are working documents of the Internet Engineering
provisions of BCP 78 and BCP 79. Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.
Internet-Drafts are working documents of the Internet Engineering Internet-Drafts are draft documents valid for a maximum of six months
Task Force (IETF). Note that other groups may also distribute and may be updated, replaced, or obsoleted by other documents at any
working documents as Internet-Drafts. The list of current Internet- time. It is inappropriate to use Internet-Drafts as reference
Drafts is at http://datatracker.ietf.org/drafts/current/. material or to cite them other than as "work in progress."
Internet-Drafts are draft documents valid for a maximum of six months This Internet-Draft will
and may be updated, replaced, or obsoleted by other documents at any expire on January 8, 2020
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 12, 2019
Copyright Notice Copyright Notice
Copyright (c) 2019 IETF Trust and the persons identified as the Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved. document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of (http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License. described in the Simplified BSD License.
Table of Contents Table of Contents
1. Introduction ................................................ 3 1. Introduction
2. Challenging Issues of IoT network ....................................................... 3
............................ 6 2. Challenging Issues of IoT network ...................................... 5
2.1. Untrusted and incorrect IoT devices ..................... 6 2.1. Untrusted and incorrect IoT devices ............................. 5
2.2. Traffic burstiness of IoT network ....................... 6 2.2. Traffic burstiness of IoT network ................................ 6
2.3. Management overheads of heterogeneous IoT sensors 2.3. Management overheads of heterogeneous IoT sensors ............... 7
........ 7 3. Overview of AI/ML-based IoT services................................... 8
3. Overview of AI/ML-based IoT services ......................... 9 4. Requirements for AI/ML-based IoT services.............................. 10
4. Requirements for AI/ML-based IoT services ................... 11 4.1. Requirements for AI/ML-based IoT data collection and delivery........ 11
4.1. Requirements for AI/ML-based IoT data collection and delivery 4.2. Requirements for intelligent and context-aware IoT services ......... 11
4.3. Requirements for applying AI/ML to IoT data ...................... 13
........................................................... 11 4.3.2. AI/ML inference in IoT application.......................... 13
4.2. Requirements for intelligent and context-aware IoT services12 5. State of arts of the artificial intelligence/machine learning technologies for IoT services14
5. State of arts of the artificial intelligence/machine learning 5.1. Machine learning and artificial intelligence technologies review ........ 14
technologies for IoT services 5.1.1. Supervised learning for IoT............................... 14
.................................. 12 5.1.2. Unsupervised learning for IoT............................. 15
5.1. Machine learning and artificial intelligence technologies 5.1.3. Reinforcement learning for IoT ............................ 16
review ..................................................... 12 5.1.4. Neural Network based algorithms for IoT..................... 17
5.1.1. Supervised learning for IoT ....................... 12 5.2. Technologies for lightweight and real-time intelligence .............. 19
5.1.2. Unsupervised learning for IoT ..................... 13 6. Use cases of AI/ML into IoT service.................................... 20
5.1.3. Reinforcement learning for IoT .................... 14 6.1. Use case 1: Surveillance and Security in Smart Home ............... 20
5.1.4. Neural Network based algorithms for IoT ........... 15 6.2. Use case 2: Smart Health in Smart Home ........................ 21
5.2. Technologies for lightweight and real-time intelligence 7. IANA Considerations ............................................... 21
. 17 8. Acknowledgements ................................................ 21
6. IANA Considerations ........................................ 18 9. Contributors ..................................................... 21
7. Acknowledgements ........................................... 18 10. Informative References ............................................ 21
8. Contributors ............................................... 18
9. Informative References
...................................... 19
1. Introduction 1. Introduction
The document explains the effects of applying artificial intelligence/machine learning (AI/M
L) algorithms in the Internet of Thing (IoT) service environments.
The document explains the effects of applying artificial IoT applications will be deployed in heterogeneous and different areas such as the energy,
intelligence/machine learning (AI/ML) algorithms in the Internet of transportation, automation and manufacturing industries as well as the information and com
Thing (IoT) service environments. munication technology (ICT) industry. Many IoT sensors and devices can connect to an IoT
service environment where IoT objects cannot interoperate with each other and can interact
with different applications. The IoT service may not run in a single administrative domain. If
market demand exists, the cross-domain service scenarios for IoT applications could be w
idely deployed. Future IoT applications occur at multiple domains of heterogeneity with vari
ous time scales.
IoT applications will be deployed in heterogeneous and different The IoT service requirements for common architectures and public APIs poses some challe
areas such as the energy, transportation, automation and nges to the underlying service environment and networking technologies. Some IoT applicat
manufacturing industries as well as the information and communication ions require signif
technology (ICT) industry. Many IoT sensors and devices can connect icant security and privacy as well as significant resource and time constra
to an IoT service environment where IoT objects cannot interoperate ints. These mission-critical applications can be separated from many common IoT applicati
with each other and can interact with different applications. The IoT
service may not run in a single administrative domain. If market
demand exists, the cross-domain service scenarios for IoT
applications could be widely deployed. Future IoT applications occur
at multiple domains of heterogeneity with various time scales.
The IoT service requirements for common architectures and public APIs icult to classify common requirements and functional requirements depending on IoT servic
poses some challenges to the underlying service environment and e scenario.
networking technologies. Some IoT applications require significant
security and privacy as well as significant resource and time
constraints. These mission-critical applications can be separated
from many common IoT applications that current technology may not
provide. It means that IoT service requirements are difficult to
classify common requirements and functional requirements depending on
IoT service scenario.
Recently, artificial intelligence technologies can help the context- Recently, artificial intelligence technologies can help the context-aware IoT service scenari
aware IoT service scenarios apply rule-based knowledge accumulation. os apply rule-based knowledge accumulation. The IoT service assumes that many sensing
The IoT service assumes that many sensing devices are connected to devices are connected to single or multiple IoT network domains. Each sensor sends small
single or multiple IoT network domains. Each sensor sends small packets to the IoT servers periodically or non-periodically. Detection data contains periodic
packets to the IoT servers periodically or non-periodically. status information that monitors whether the system is in a normal state or not. In some ca
ses, alert information is included for quick processing. Most IoT applications can operate in
two modes. One is a simple monitoring mode and the other is an abnormal mode for rapid
processing. In a simple monitoring phase, the IoT device periodically sends sensing data to
the server. If the measured data is outside the normal range, the IoT service can change th
e operating mode to an abnormal phase and activate future probes. Alarm conditions shoul
d be promptly notified to responsible persons. For mission-critical applications, reliable co
mmunication with robust QoS requirements in terms of error and latency is required.
Detection data contains periodic status information that monitors Periodic data accumulation from IoT devices is cumbersome. Under normal conditions, the
whether the system is in a normal state or not. In some cases, alert IoT data is simply accumulated without further action. In an unusual situation, incoming IoT
information is included for quick processing. Most IoT applications data can cause an urgent action to notify the administrator of the problem. Streaming data t
can operate in two modes. One is a simple monitoring mode and the raffic from thousands of IoT devices is annoying to store in the database because it is not
other is an abnormal mode for rapid processing. In a simple easy to extract unidentified or future incidents. Only a significant portion of the incoming da
monitoring phase, the IoT device periodically sends sensing data to ta stream can be stored in a real-time database that is time-sensitive and capable of rapid
the server. If the measured data is outside the normal range, the IoT query processing. A combination of different IoT detection data, including location, time, a
service can change the operating mode to an abnormal phase and nd status, allows you to sort and categorize a portion of streaming data when an additional
activate future probes. Alarm conditions should be promptly notified inspection is required, and perform real-time processing. One of the missions of the IoT da
to responsible persons. For mission-critical applications, reliable tabase is to be able to extract preliminary symptoms of unexpected accidents from a large
communication with robust QoS requirements in terms of error and amount of streaming data.
latency is required.
Periodic data accumulation from IoT devices is cumbersome. Under If some transmitted data is important to invoke the corresponding action, there are some qu
normal conditions, the IoT data is simply accumulated without further estions about whether the incoming data is correct. If the incoming data contains accurate
action. In an unusual situation, incoming IoT data can cause an and time-critical events, appropriate real-time control and management can be performed.
urgent action to notify the administrator of the problem. Streaming
data traffic from thousands of IoT devices is annoying to store in
the database because it is not easy to extract unidentified or future
incidents. Only a significant portion of the incoming data stream can
be stored in a real-time database that is time-sensitive and capable
of rapid query processing. A combination of different IoT detection
data, including location, time, and status, allows you to sort and
categorize a portion of streaming data when an additional inspection
is required, and perform real-time processing. One of the missions of
the IoT database is to be able to extract preliminary symptoms of
unexpected accidents from a large amount of streaming data.
If some transmitted data is important to invoke the corresponding may occur. In these cases, incoming data can trigger to initiate additional inspections to pr
action, there are some questions about whether the incoming data is otect against future unacceptable situations. But, if time-critical data is missed due to error
correct. If the incoming data contains accurate and time-critical s in the sensing devices and the delivery protocol, there is no reason to configure IoT netw
events, appropriate real-time control and management can be performed. orks and devices at a high cost.
However, if the incoming data is inaccurate or intentionally
corrupted, additional accidents may occur. In these cases, incoming
data can trigger to initiate additional inspections to protect
against future unacceptable situations. But, if time-critical data is
missed due to errors in the sensing devices and the delivery protocol,
there is no reason to configure IoT networks and devices at a high
cost.
It is not easy to analyze data collected through IoT devices It is not easy to analyze data collected through IoT devices installed to monitor complex IoT
installed to monitor complex IoT service environments. If the sensor service environments. If the sensor malfunctions, the data of the sensor cannot be trusted.
malfunctions, the data of the sensor cannot be trusted. Additional Additional investigation should be done if abnormal status from specific sensors is collecte
investigation should be done if abnormal status from specific sensors
is collected. The data of the redundant sensor installed in the same
area should be received or combined with other sensor information
adjacent to the sensor to determine the abnormal state.
For sensors installed in a specific area, sensing records will remain d. The data of the redundant sensor installed in the same area should be received or combi
for a certain period of time. IoT service operators can look at the ned with other sensor information adjacent to the sensor to determine the abnormal state.
operational history of the sensor for a period of time to determine
what problems were encountered when data was collected. When an
abnormal situation occurs, IoT sensor should investigate whether it
noticed normal operations and notified the IoT service operator. If
the abnormal situation is not properly detected, the operator should
analyze whether it was caused by malfunction of the IoT sensor or
other reasons.
In the IoT service environment, it is possible to analyze the For sensors installed in a specific area, sensing records will
situation accurately by applying recent artificial intelligence and remain for a certain period of ti
machine learning technologies. If there is an operational record of me. IoT service operators can look at the operational history of the sensor for a period of ti
the past, it is possible to determine when an abnormal situation me to determine what problems were encountered when data was collected. When an abno
arises. Most problems are likely to be repeated, so if the past rmal situation occurs, IoT sensor should investigate whether it noticed normal operations an
learning experience is accumulated, the anomaly of IoT services can d notified the IoT service operator. If the abnormal situation is not properly detected, the op
be easily and immediately identified. In addition, when information erator should analyze whether it was caused by malfunction of the IoT sensor or other reas
gathered from various sensors is synthesized, it is possible to ons.
accurately determine whether abnormal situations have occurred.
Various types of IoT sensors are installed with certain purposes. It In the IoT service environment, it is possible to analyze the situation accurately by applying
expects that all the IoT sensors intend to monitor the occurrence of recent artificial intelligence and machine learning technologies. If there is an operational rec
special abnormal situations in advance. Therefore, it should be set ord of the past, it is possible to determine when an abnormal situation arises. Most problem
in advance what actions are required when a specific anomaly occurs. s are likely to be repeated, so if the past learning experience is accumulated, the anomaly o
The appropriate work is performed on the abnormal situation according f IoT services can be easily and immediately identified. In addition, when information gather
to the procedure, predefined by the human. By using artificial ed from various sensors is synthesized, it is possible to accurately determine whether abnor
intelligence and machine learning algorithms, the appropriate actions mal situations have occurred.
are taken when an abnormal situation is detected from various IoT
sensors.
2. Challenging Issues of IoT network Various types of IoT sensors are installed with certain purposes. It expects that all the IoT s
ensors intend to monitor the occurrence of special abnormal situations in advance. Therefo
re, it should be set in advance what actions are required when a specific anomaly occurs. T
he appropriate work is performed on the abnormal situation according to the procedure, pre
This section describes the challenging issues of data sensing, ppropriate actions are taken when an abnormal situation is detected from various IoT senso
collection, transfer, and intelligent decision from untrusted data rs.
quality and unexpected situations of IoT service environments.
2.1. Untrusted and incorrect IoT devices 2. Challenging Issues of IoT network
This section describes the challenging issues of data sensing, collection, transfer, and
intelligent decision from untrusted data quality and unexpected situations of IoT service
environments.
IoT traffic is similar to traditional Internet traffic with small 2.1. Untrusted and incorrect IoT devices
packet sizes. Mobile IoT traffic can cause some errors and delays IoT traffic is similar to traditional Internet traffic with small packet sizes. Mobile IoT
because wireless links are unstable and signal strength may be traffic can cause some errors and delays because wireless links are unstable and signal
degraded with device mobility. If the signal strength of the IoT strength may be degraded with device mobility. If the signal strength of the IoT device
device with a power limit is not so strong, the reception quality of with a power limit is not so strong, the reception quality of the IoT server may not be
the IoT server may not be sufficient to obtain the measurement data. sufficient to obtain the measurement data.
For mission-critical applications, such as smart-grid and factory- For mission-critical applications, such as smart-grid and factory-automation,
automation, expensive IoT sensors with self-rechargeable batteries expensive IoT sensors with self-rechargeable batteries and redundant hardware logic
and redundant hardware logic may be required. However, unexpected may be required. However, unexpected abnormal situations may occur due to sensor
abnormal situations may occur due to sensor malfunctions. There are malfunctions. There are trade-offs between implementation cost and efficiency for
trade-offs between implementation cost and efficiency for cost- cost-effective IoT services. When smart-grid and factory-automation applications are
effective IoT services. When smart-grid and factory-automation equipped with IoT devices, the acceptable quality from IoT solutions can be required.
applications are equipped with IoT devices, the acceptable quality Sometimes, expensive and duplicated IoT solutions may be needed.
from IoT solutions can be required. Sometimes, expensive and
duplicated IoT solutions may be needed.
2.2. Traffic burstiness of IoT network 2.2. Traffic burstiness of IoT network
IoT traffic includes two types of traffic characteristic: periodic with small packet sizes
and bursty with high bandwidth. Under normal conditions, the IoT traffic periodically
transmits status information with a small bandwidth, several kilobits/sec. However, in an
abnormal state, IoT devices need a high bandwidth, up to several tens of megabits/sec,
in order to identify actual events and investigate accurate status information. In additi
on,
traffic volume can explosively increase in response to emergencies. For example, in the
case of smart-grid application, the bandwidth of several kilobits/sec is usually used,
and when an urgent situation occurs, a broadband channel is required up to several
tens of megabits/sec.
IoT traffic includes two types of traffic characteristic: periodic The other traffic can be integrated at an IoT network to increase bandwidth efficiency. If
with small packet sizes and bursty with high bandwidth. Under normal an emergency situation occurs in the IoT service, IoT traffic volumes suddenly increase,
conditions, the IoT traffic periodically transmits status information in which case network processing capacity may be not sufficient. If the IoT service is
with a small bandwidth, several kilobits/sec. However, in an abnormal integrated with voice and video applications, the problem can become more complex.
state, IoT devices need a high bandwidth, up to several tens of As time goes by, traffic congestion and bottlenecks are frequent in some areas. In
megabits/sec, in order to identify actual events and investigate addition, if an existing service policy changes (for example, prioritizing certain traffic or
accurate status information. In addition, traffic volume can suddenly changing the route), other unexpected problems may be encountered. Various
explosively increase in response to emergencies. For example, in the congestion control and load balancing algorithms with the help of artificial intell
case of smart-grid application, the bandwidth of several kilobits/sec Until now, much research has been done on traffic variability in an integrated network
is usually used, and when an urgent situation occurs, a broadband service environment. All networks have their own traffic characteristics, depending on
channel is required up to several tens of megabits/sec. geographical area, number of subscribers, subscribers' preferences, and types of
applications used. In the case of IoT traffic, the normal bandwidth is very small. If the
The other traffic can be integrated at an IoT network to increase IoT traffic volume increases abruptly in an abnormal situation, the network may suffer
bandwidth efficiency. If an emergency situation occurs in the IoT unacceptable delay and loss. If emergency situations detected by IoT networks occur in
service, IoT traffic volumes suddenly increase, in which case network a smart grid or intelligent transportation system, the processing power of the IoT
processing capacity may be not sufficient. If the IoT service is network alone cannot solve the problem and the help of existing network resources is
integrated with voice and video applications, the problem can become inevitable.
more complex. As time goes by, traffic congestion and bottlenecks are
frequent in some areas. In addition, if an existing service policy
changes (for example, prioritizing certain traffic or suddenly
changing the route), other unexpected problems may be encountered.
Various congestion control and load balancing algorithms with the
help of artificial intelligence can be applied to handle time-varying
traffic on a network.
Until now, much research has been done on traffic variability in an
integrated network service environment. All networks have their own
traffic characteristics, depending on geographical area, number of
subscribers, subscribers' preferences, and types of applications used.
In the case of IoT traffic, the normal bandwidth is very small. If
the IoT traffic volume increases abruptly in an abnormal situation,
the network may suffer unacceptable delay and loss. If emergency
situations detected by IoT networks occur in a smart grid or
intelligent transportation system, the processing power of the IoT
network alone cannot solve the problem and the help of existing
network resources is inevitable.
2.3. Management overheads of heterogeneous IoT sensors 2.3. Management overheads of heterogeneous IoT sensors
Traffic management in an integrated network environment is not easy. In order to
operate the network steadily, a network operator has its own know-hows and
experiences. If there are plenty of network resources, it is easy to set up a bypass route
even if network failure or congestion occurs in a specific area. For operating network
steadily, network resources may be designed to be over-provisioned in order to cope
with various possible outages. A network operator predicts the amount of traffic
generated by the corresponding equipment and grasps to what extent a transmission
bandwidth is required. If traffic fluctuation is very severe, the network operator can
allocate network resources in advance. In case of frequent failures or severe traffic
fluctuation, some network resources are separated in order not to affect normal traffic.
Traffic management in an integrated network environment is not easy. More than a billion IoT devices are expected to connect to smartphones, tablets,
In order to operate the network steadily, a network operator has its wearables, and vehicles. Therefore, IoT services are targeted at mobile applications. In
own know-hows and experiences. If there are plenty of network particular, intelligent transportation systems need the help of IoT technology to provide
resources, it is easy to set up a bypass route even if network traffic monitoring and prevent public or private traffic accidents. IoT technology can
failure or congestion occurs in a specific area. For operating play an important role in reducing traffic congestion, saving people's travel time and
network steadily, network resources may be designed to be over- The IoT service has troublesome administrative problems to configure an IoT network
provisioned in order to cope with various possible outages. A network which consists of IoT servers, gateways, and many sensing devices. The small-sized
operator predicts the amount of traffic generated by the but large-numbered IoT devices may incur administrative overhead since all the IoT
corresponding equipment and grasps to what extent a transmission devices should be initialized and the bootstrapping information of IoT resources should
bandwidth is required. If traffic fluctuation is very severe, the be loaded into the IoT service environments. Whenever some IoT devices are newly
network operator can allocate network resources in advance. In case added and some devices have to be removed, the dynamic reconfiguration of IoT
of frequent failures or severe traffic fluctuation, some network resources is essential. In addition, the IoT device's preinstalled software should be
resources are separated in order not to affect normal traffic. regularly inspected and upgraded according to its version. Frequent upgrades and
changes to some IoT devices may require autonomic management and bootstrapping
More than a billion IoT devices are expected to connect to techniques.
smartphones, tablets, wearables, and vehicles. Therefore, IoT
services are targeted at mobile applications. In particular,
intelligent transportation systems need the help of IoT technology to
provide traffic monitoring and prevent public or private traffic
accidents. IoT technology can play an important role in reducing
traffic congestion, saving people's travel time and costs, and
providing a pleasant journey.
The IoT service has troublesome administrative problems to configure
an IoT network which consists of IoT servers, gateways, and many
sensing devices. The small-sized but large-numbered IoT devices may
incur administrative overhead since all the IoT devices should be
initialized and the bootstrapping information of IoT resources should
be loaded into the IoT service environments. Whenever some IoT
devices are newly added and some devices have to be removed, the
dynamic reconfiguration of IoT resources is essential. In addition,
the IoT device's preinstalled software should be regularly inspected
and upgraded according to its version. Frequent upgrades and changes
to some IoT devices may require autonomic management and
bootstrapping techniques.
Network management generally assumes that all network resources Network management generally assumes that all network resources operate reliably with
operate reliably with acceptable quality. In most failure situations, acceptable quality. In most failure situations, the network operator decides to switch to
the network operator decides to switch to a redundant backup device a redundant backup device or bypass the failed communication path. If some IoT
or bypass the failed communication path. If some IoT devices are not devices are not stable, duplicate IoT devices can be installed for the same purpose. If
stable, duplicate IoT devices can be installed for the same purpose. IoT resources are not duplicated, various mechanisms are needed to reduce the
If IoT resources are not duplicated, various mechanisms are needed to damage. Therefore, it is necessary to prioritize the management tasks to be performed
reduce the damage. Therefore, it is necessary to prioritize the first when an abnormality occurs in the IoT service environment. However, managing
management tasks to be performed first when an abnormality occurs in duplicate networks can cause another problem. If two IoT devices are running at the
the IoT service environment. However, managing duplicate networks can same time, the recipient can get redundant information. If two or more unusual
cause another problem. If two IoT devices are running at the same situations occur at the same time, it is difficult to solve the problem since tasks for
time, the recipient can get redundant information. If two or more urgent processing should be distinguished from tasks that can be performed over time.
unusual situations occur at the same time, it is difficult to solve
the problem since tasks for urgent processing should be distinguished
from tasks that can be performed over time.
In addition, the operations manager's mistakes or misunderstanding of In addition, the operations manager's mistakes or misunderstanding of problem
problem situations can lead to other unexpected complications. situations can lead to other unexpected complications. Therefore, artificial intelligence
Therefore, artificial intelligence technologies can help what kind of technologies can help what kind of network management work is required when an
network management work is required when an unexpected complicated unexpected complicated situation occurs even though a procedure for an abnormal
situation occurs even though a procedure for an abnormal situation is situation is already prepared.
already prepared.
3. Overview of AI/ML-based IoT services 3. Overview of AI/ML-based IoT services
In this section, successful applications of artificial intelligence in IoT domains are provided.
In this section, successful applications of artificial intelligence her than later analytics with piled data. Recently, neural-network-based artificial intelligence
in IoT domains are provided. The common property of IoT applications technologies are widely used across many IoT applications.
and services is that they require fast analytics rather than later
analytics with piled data. Recently, neural-network-based artificial
intelligence technologies are widely used across many IoT
applications.
Simple IoT applications include dynamic contexts that share common Simple IoT applications include dynamic contexts that share common features among socia
features among social relations at the same administration domain. l relations at the same administration domain. IoT devices in the same domain can provide t
IoT devices in the same domain can provide their service contexts to heir service contexts to the IoT server. When a dynamic change occurs in an IoT service co
the IoT server. When a dynamic change occurs in an IoT service ntext, the IoT device needs real-time processing to activate urgent events, alert notification
context, the IoT device needs real-time processing to activate urgent s, update, and reconnect contexts. The IoT service must support real-time interactions bet
events, alert notifications, update, and reconnect contexts. The IoT ween the IoT device and the system in the same domain. The IoT service contexts must be
service must support real-time interactions between the IoT device shared between physical objects and social members in the same domain as well.
and the system in the same domain. The IoT service contexts must be
shared between physical objects and social members in the same domain
as well.
Artificial intelligence technologies have been shown promising in Artificial intelligence technologies have been shown promising in many areas, including IoT.
many areas, including IoT. For example, contextual information for a For example, contextual information for a car-sharing business must interact with custome
car-sharing business must interact with customers, car owners, and rs, car owners, and car sharing providers. All entities in the value chain of a car sharing bus
car sharing providers. All entities in the value chain of a car iness must share the corresponding situation to pick up, board, and return shared cars. Co
sharing business must share the corresponding situation to pick up, mmunication networks and interactive information, including registration and payment, can
board, and return shared cars. Communication networks and interactive be shared tightly among the entities. Home IoT service environment can be equipped with s
information, including registration and payment, can be shared ensors for theft detection, door lock, temperature, fire detection, gas detection, short circui
tightly among the entities. Home IoT service environment can be t, air condition to name a few. Office IoT service environments,
equipped with sensors for theft detection, door lock, temperature, including buildings such as
fire detection, gas detection, short circuit, air condition to name a shopping centers and bus/airport terminals, have their own sensors, including alarm sensor
few. Office IoT service environments, including buildings such as s. When an alarm signal is detected by the sensor, the physical position and occurrence ti
shopping centers and bus/airport terminals, have their own sensors, me of the sensor is determined in advance. All signals from various sensors are analyzed c
including alarm sensors. When an alarm signal is detected by the omprehensively to make the right decision. If some sensors frequently malfunction, the situ
sensor, the physical position and occurrence time of the sensor is ation can be grasped more accurately by analyzing the information of the adjacent sensor. I
determined in advance. All signals from various sensors are analyzed n particular, when installing multiple sensors in a particular building (e.g., surveillance came
comprehensively to make the right decision. If some sensors ra, location monitoring, temperature, etc.), a much wider range of sensors can be used wh
frequently malfunction, the situation can be grasped more accurately en utilizing artificial intelligence and machine learning technologies.
by analyzing the information of the adjacent sensor. In particular,
when installing multiple sensors in a particular building (e.g.,
surveillance camera, location monitoring, temperature, etc.), a much
wider range of sensors can be used when utilizing artificial
intelligence and machine learning technologies.
(Smart home) Smart home concept span over multiple IoT applications, (Smart home) Smart home concept span over multiple IoT applications, health, energy, ente
health, energy, entertainment, education, etc. It involves voice rtainment, education, etc. It involves voice recognition, natural language processing, image
recognition, natural language processing, image-based object -based object recognition, appliance management, and many more artif
recognition, appliance management, and many more artificial icial intelligence te
intelligence technologies integrated with IoT. Smart connected- er control over home supplies and expenses. The energy consumption and efficiency of ho
devices monitor the house to provide better control over home me appliances are monitored and analyzed with deep learning based technologies, such as
supplies and expenses. The energy consumption and efficiency of home artificial neural network, long-short-term-memory, etc.
appliances are monitored and analyzed with deep learning based
technologies, such as artificial neural network, long-short-term-
memory, etc.
(Smart city) Smart city, as well, contains multiple IoT domains, (Smart city) Smart city, as well, contains multiple IoT domains, transportation, infrastructure,
transportation, infrastructure, energy, agriculture, etc. Since energy, agriculture, etc. Since heterogeneous data from different domains are gathered in
heterogeneous data from different domains are gathered in smart smart cities, various artificial intelligence approaches are studied in smart-city application.
cities, various artificial intelligence approaches are studied in Public transportation behaviors and crowd movements patterns are important issues, and th
smart-city application. Public transportation behaviors and crowd ey are often dealt with neural network based methods, long-short-term-memory and convo
movements patterns are important issues, and they are often dealt lutional neural network.
with neural network based methods, long-short-term-memory and
convolutional neural network.
(Smart energy) As two-way communication energy infrastructure is (Smart energy) As two-way communication energy infrastructure is deployed, smart grid ha
deployed, smart grid has become a big IoT application, which requires s become a big IoT application, which requires intelligent data processing. The traditional e
intelligent data processing. The traditional energy providers are nergy providers are highly interested in recognizing local energy consumption patterns and
highly interested in recognizing local energy consumption patterns forecasting the needs in order to make appropriate decisions on real-time. Moreover, the e
and forecasting the needs in order to make appropriate decisions on nergy consumers, as well, want analyzed information on their own energy consumption beh
real-time. Moreover, the energy consumers, as well, want analyzed aviors. Recently, many works on energy consumption prediction, energy flexibility analysis,
information on their own energy consumption behaviors. Recently, many etc. are actively ongoing. Most works are based on the latest deep learning technologies, s
works on energy consumption prediction, energy flexibility analysis, uch as multi-layered-perceptron, recurrent neural network, long-short-term-memory, auto
etc. are actively ongoing. Most works are based on the latest deep encoder, etc.
learning technologies, such as multi-layered-perceptron, recurrent
neural network, long-short-term-memory, autoencoder, etc.
(Smart transportation) The intelligent transportation system is (Smart transportation) The intelligent transportation system is another source of big data in
another source of big data in IoT domains. Many use cases, such as IoT domains. Many use cases, such as traffic flow and congestion prediction, traffic sign re
traffic flow and congestion prediction, traffic sign recognition, cognition, vehicle intrusion detection, etc., have been studied. Moreover, a lot of advanced
vehicle intrusion detection, etc., have been studied. Moreover, a lot artificial intelligence technologies are required in autonomous and smart vehicles, which re
of advanced artificial intelligence technologies are required in quire many intelligent sub-tasks, such as pedestrian's detection, obstacle avoidance, etc.
autonomous and smart vehicles, which require many intelligent sub-
tasks, such as pedestrian's detection, obstacle avoidance, etc.
(Smart healthcare) IoT and artificial intelligence are integrated (Smart healthcare) IoT and artificial intelligence are integrated into the healthcare and wellb
into the healthcare and wellbeing domain as well. By analyzing food eing domain as well. By analyzing food images with convolutional neural network on mobile
images with convolutional neural network on mobile devices, dietary devices, dietary intakes can be measured. With voice signal captured from sensor devices,
intakes can be measured. With voice signal captured from sensor voice pathologies can be detected. Moreover, recurrent neural network and long-short-ter
devices, voice pathologies can be detected. Moreover, recurrent m-memory technologies are actively being studied for early diagnosis and prediction of dis
neural network and long-short-term-memory technologies are actively eases with time series medical data.
being studied for early diagnosis and prediction of diseases with
time series medical data.
(Smart agriculture) To manage a vast area of land, IoT and artificial (Smart agriculture) To manage a vast area of land, IoT and artificial intel ligence technologie
intelligence technologies are recently used in agriculture domains. twork are utilized for crop detection or classification and disease recognition in the plants.
Deep neural network and convolutional neural network are utilized for Moreover, for automatic farming with autonomous machine operation, obstacle avoidance,
crop detection or classification and disease recognition in the fruit location, and many more sub-tasks are handled with advanced artificial intelligence te
plants. Moreover, for automatic farming with autonomous machine chnologies.
operation, obstacle avoidance, fruit location, and many more sub-
tasks are handled with advanced artificial intelligence technologies.
4. Requirements for AI/ML-based IoT services 4. Requirements for AI/ML-based IoT services
In this section, the requirements for AI/ML-based IoT data collection and delivery, intelligen
(to be included) t and context-aware IoT services, and applying AI/ML to IoT data willbe described.
4.1. Requirements for AI/ML-based IoT data collection and delivery 4.1. Requirements for AI/ML-based IoT data collection and delivery
IoT services store a vast amount of data that IoT devices periodically generate, and the
refining and analyzing are costly. Effective analysis of IoT data has been considered to
be the most important factor in data processing, but the analysis of efficient data
collection and delivery methods are becoming other significant factors as the amount of
the data collected is explosively increasing.
(to be included) In particular, as a number of IoT devices have been deployed within the IoT network,
controlling data collection and delivery for each of them has become impossible. The
introduction of AI/ML techniques for simultaneous and efficient management of the IoT
devices should be considered as a countermeasure. For IoT data collection and
delivery, the following two factors will need to be considered, IoT devices energy and
data quality.
(IoT Device Energy) As many IoT devices have begun to be deployed within the IoT
network, it is impossible to deliver energy to many IoT devices simultaneously.
Consequently, the efficient battery use has become an important issue.
If IoT data collection and delivery periods are too short, a lifetime of the IoT device will
be shortened through the reckless use of IoT device energy. Thereby, it increases the
cost required to provide IoT service. On the other hand, if IoT data collection and
delivery period are too long, the quality of the IoT services provided will be reduced due
to the lack of details in the data for situation recognition and real-time processing.
Therefore, taking into account the energy consumption of the IoT devices, research on
proper IoT data collection and delivery period is necessary.
(Data Quality) Since the data collected from the majority of IoT devices usually contain
redundant information, it causes additionalcosts for the data collection and refinement
processes. Therefore, it will be necessary to select and deliver meaningful information
from redundant IoT data to reduce unnecessary cost on the IoT network. To do so, it
will need the research to identify the relationships among the data collected various
4.2. Requirements for intelligent and context-aware IoT services 4.2. Requirements for intelligent and context-aware IoT services
In a context-aware IoT service environment, it is important to establish a context to be
aware of in advance since IoT devices will be deployed according to a pre-designed
architecture and to check how characteristics of IoT data and data-to-data
characteristics are expressed under these circumstances. For the data produced by IoT
devices, since it contains the device's relative location information, sensing value over
time, event information, it should be reviewed to provide the target context-aware
service using this information. Some of the necessary technologies will be described in
the following.
(to be included) (Physical Clustering) To increase the accuracy of context-awareness, the provision of
context-aware services should be considered in a situation where the relationship
between IoT devices with respect to physical layout or physical environment is taken
into account. Setting a rule using the service provider's domain knowledge may be
possible, but introducing the physical clustering into a diverse IoT environment(e.g., in
bedroom, kitchen, balcony, or a space connected through an open door) will require
identifying the physical relationship between the devices using data generated from IoT
devices.
5. State of arts of the artificial intelligence/machine learning (Extra Data Processing) In order to prevent degradation of service quality from errors in
technologies for IoT services data values or device malfunctions, extra sensors should be placed in the majority of
IoT environments. In a context-aware service, they contain the same information, so
the technologies filtering the data that contains only essential part among the same
information while preventing data errors would be required.
In this section, well-known machine learning and artificial (Unreported data handling) If an event is detected on a particular IoT device, it will
intelligence technologies applicable to IoT applications are reviewed.
transmit data regardless of the device's sensing and delivery interval. At this time, the
data of IoT devices which are physically clustered are needed to accurately detect
events that occurred, and it is difficult to expect that these devices will provide data at
problem for real-time processing for emergency situations. Therefore, handling
unreported data will be required based on previously collected data.
(Abnormal data in AI/ML) In the case of context-aware services that operates based on
the predetermined rule, the flexibility to cope with emergency situations that have not
been considered is low, and thus AI/ML algorithms are required to intelligently cope
with a myriad of situations. However, many abnormal data are generated depending on
environmental conditions such as device status, so AI/ML algorithms that can operate
in that imperfect environment should be considered.
4.3. Requirements for applying AI/ML to IoT data
In this subsection, the requirements for applying AI/ML to IoT data are described.
4.3.1. Training AI/ML algorithm
To use AI/ML algorithm, two elements are required, AI/ML model and training data. The
presence of training dataset in good quality is an important factor of the AI/ML model
performance since the model is iteratively trained with the training data. However, for
anomaly detection, there is not enough training data since not only the probability of
anomaly occurrence is very low but also it is almost impossible to retrieve the ground
truth value even when the situation has occurred. Therefore, using domain knowledge,
AI/ML learning based on abnormal situation data generation or simulation should be
considered. For example, for an external intrusion detection application within a smart
home, when a camera and a motion sensor detect an intruder, a light sensor checks
the measuring value. If the light does not turn on, then the IoT application recognizes it
as an abnormal situation. In this way, by using the domain knowledge, the rule
regarding the operational scenario of the IoT application is generated as the training
data, and the generated training data can be used for model learning. This will not only
enable learning the anomaly detection algorithm in IoT application but also improving
the accuracy. Therefore, IoT application, in which it is difficult to acquire dataset in
good quality, willrequire data generation based on domain knowledge for AI/ML.
4.3.2. AI/ML inference in IoT application
In order for AI ML technology to be applied to IoT applications, the training data and the
input data for model testing and inferencing must have the same characteristics such
as dimension, time interval, types of features, etc. However, due to the volatile IoT data
characteristics that vary from situations in many IoT applications, it is difficult to directly
periodically send sensing data, and AI/ML have no difficulty in operating. However, in
an abnormal mode, the IoT applications require a fast response, and IoT devices
transmit data at shorter intervals than normal, which changes the characteristics of the
data being input to the AI/ML algorithm. Therefore, data preprocessing technology
handling the abnormal data will be required in advance, such as data imputation,
correcting data anomalies, and Interpolation of unreported data.
5. State of arts of the artificial intelligence/machine learning technologies for IoT services
In this section, well-known machine learning and artificial intelligence technologies
applicable to IoT applications are reviewed.
5.1. Machine learning and artificial intelligence technologies review 5.1. Machine learning and artificial intelligence technologies review
The classical machine learning models can be divided into three types, The classical machine learning models can be divided into three types, supervised,
supervised, unsupervised, and reinforcement learnings. Therefore, in unsupervised, and reinforcement learnings. Therefore, in this subsection, machine
this subsection, machine learning and artificial intelligence learning and artificial intelligence technology reviews are done in four different
technology reviews are done in four different categories: supervised, categories: supervised, unsupervised, reinforcement, and neural-network-based.
unsupervised, reinforcement, and neural-network-based.
5.1.1. Supervised learning for IoT 5.1.1. Supervised learning for IoT
Supervised learning is a task-based type of machine learning, which Supervised learning is a task-based type of machine learning, which approximates
approximates function describing the relationship and causality function describing the relationship and causality between input and output data.
between input and output data. Therefore, the input data needs to be Therefore, the input data needs to be clearly defined with proper output data since
clearly defined with proper output data since supervised learning supervised learning models learn explicitly from direct feedback.
models learn explicitly from direct feedback.
(K-Nearest Neighbor) Given a new data point in K-Nearest Neighbor (K-Nearest Neighbor) Given a new data point in K-Nearest Neighbor (KNN) classifier, it
(KNN) classifier, it is classified according to its K number of the is classified according to its K number of the closest data points in the training set. To
closest data points in the training set. To find the K nearest find the K nearest neighbors of the new data point, it needs to use a distance metric
neighbors of the new data point, it needs to use a distance metric which can affect classifier performance, such as Euclidean, Mahalanobis or Hamming.
which can affect classifier performance, such as Euclidean, One limitation of KNN in applying for IoT network is that it is unscalable to large
Mahalanobis or Hamming. One limitation of KNN in applying for IoT datasets because it requires the entire training dataset to classify a newly incoming
network is that it is unscalable to large datasets because it data. However, KNN required less processing power capability compared to other
requires the entire training dataset to classify a newly incoming complex learning methods.
data. However, KNN required less processing power capability compared
to other complex learning methods.
(Naive Bayes) Given a new data point in Naive Bayes classifiers, (Naive Bayes) Given a new data point in Naive Bayes classifiers,
it is classified based on Bayes' theorem with the "naive" assumption it is classified based on Bayes' theorem with the "naive" assumption of independence
of independence between the features. Since Naive Bayes classifiers between the features. Since Naive Bayes classifiers don't need a large number of data
don't need a large number of data points to be trained, they can deal points to be trained, they can deal with high-dimensional data points. Therefore, they
with high-dimensional data points. Therefore, they are fast and are fast and highly scalable. However, since its "naive" assumptions are somewhat
highly scalable. However, since its "naive" assumptions are somewhat
strong, a certain level of prior knowledge on the dataset is required. strong, a certain level of prior knowledge on the dataset is required.
(Support Vector Machine) Support Vector Machine (SVM) is a binary and (Support Vector Machine) Support Vector Machine (SVM) is a binary and non-
non-probabilistic classifier which finds the hyperplane maximizing probabilistic classifier which finds the hyperplane maximizing the margin between the
the margin between the classes of the training dataset. SVM has been classes of the training dataset. SVM has been the most pervasive machine learning
the most pervasive machine learning technology until the study on technology until the study on neural network technologies are advanced recently.
neural network technologies are advanced recently. However, SVM still However, SVM still has advantages over neural network based and probabilistic
has advantages over neural network based and probabilistic approaches approaches in terms of memory usage and capability to deal with high-dimensional
in terms of memory usage and capability to deal with high-dimensional data. In this manner, SVM can be used for IoT applications with severe data storage
data. In this manner, SVM can be used for IoT applications with constraint.
severe data storage constraint.
(Regression) Regression is a method for approximating the (Regression) Regression is a method for approximating the relationships of the
relationships of the dependent variable, which is being estimated, dependent variable, which is being estimated, with the independent variables, which are
with the independent variables, which are used for the estimation. used for the estimation. Therefore, this method is widely used for forecasting and
Therefore, this method is widely used for forecasting and inferring inferring causal relationships between input data and output data in time-sensitive IoT
causal relationships between input data and output data in time- application.
sensitive IoT application.
(Random Forests) In random forests, instead of training a single (Random Forests) In random forests, instead of training a single decision tree, a group
decision tree, a group of trees is trained. Each tree is trained on a of trees is trained. Each tree is trained on a subset of the training set using a randomly
subset of the training set using a randomly chosen subset of M input chosen subset of M input variables. Random forests considering various tree structures
variables. Random forests considering various tree structures have have very high accuracy, so it can be utilized in the accuracy-critical IoT applications.
very high accuracy, so it can be utilized in the accuracy-critical
IoT applications.
5.1.2. Unsupervised learning for IoT 5.1.2. Unsupervised learning for IoT
Unsupervised learning is a data-driven type of machine learning which Unsupervised learning is a data-driven type of machine learning which finds hidden
finds hidden structure in unlabeled dataset without feedback during structure in unlabeled dataset without feedback during the learning process. Unlike
the learning process. Unlike supervised learning, unsupervised supervised learning, unsupervised learning focuses on discovering patterns in the data
learning focuses on discovering patterns in the data distributions distributions and gaining insights from them.
and gaining insights from them.
(K-means clustering) K-means clustering aims to assign observations (K-means clustering) K-means clustering aims to assign observations into K number of
into K number of clusters in which each observation belongs to the clusters in which each observation belongs to the cluster having the most similarities.
cluster having the most similarities. The measure of similarity is The measure of similarity is the distance between K cluster centers and each
the distance between K cluster centers and each observation. K-means observation. K-means is a very fast and highly scalable clustering algorithm, so it can
is a very fast and highly scalable clustering algorithm, so it can be be used for IoT applications with real-time processing requirements such as smart
used for IoT applications with real-time processing requirements such transportation.
as smart transportation.
(Density-based spatial clustering of applications with noise) (Density-based spatial clustering of applications with noise)
Density-Based approach to Spatial Clustering of Applications with Density-Based approach to Spatial Clustering of Applications with Noise (DBSCAN) is a
Noise (DBSCAN) is a method that clusters dataset based on the density method that clusters dataset based on the density of its data samples. In this model,
of its data samples. In this model, dense regions which include data dense regions which include data samples with many close neighbors are considered
samples with many close neighbors are considered as clusters, and as clusters, and data samples in low-density regions are classified as outliers [Kriegal].
data samples in low-density regions are classified as outliers Since this method is robust to outliers, DBSCAN is efficient data clustering method for
[Kriegal]. Since this method is robust to outliers, DBSCAN is IoT network environments with untrusted big datasets in practice.
efficient data clustering method for IoT network environments with
untrusted big datasets in practice.
5.1.3. Reinforcement learning for IoT 5.1.3. Reinforcement learning for IoT
Reinforcement learning is a reactive type of machine learning that Reinforcement learning is a reactive type of machine learning that learn a series of
learn a series of actions in a given set of possible states, actions, actions in a given set of possible states, actions, and rewards or penalties. It can be
and rewards or penalties. It can be seen as the exploring decision- seen as the exploring decision-making process and choosing the action series with the
making process and choosing the action series with the most reward or most reward or the least penalty which can be cost, priority, time to name a few.
the least penalty which can be cost, priority, time to name a few. Reinforcement learning can be helpful for selecting action of IoT device by providing a
Reinforcement learning can be helpful for selecting action of IoT guideline.
device by providing a guideline.
(Q-learning) Q-Learning is a model-free, off-policy reinforcement (Q-learning) Q-Learning is a model-free, off-policy reinforcement learning algorithm
learning algorithm based on the well-known Bellman Equation. The goal based on the well-known Bellman Equation. The goal is to learn an action-selection
is to learn an action-selection policy maximizing the Q-value, which policy maximizing the Q-value, which tells an agent what action to take. It can be used
tells an agent what action to take. It can be used for IoT device to for IoT device to determine which action it should take according to conditions.
determine which action it should take according to conditions.
(State-Action-Reward-State-Action) Though State-Action-Reward-State- (State-Action-Reward-State-Action) Though State-Action-Reward-State-Action
Action (SARSA) is a much similar algorithm to Q-learning, the main (SARSA) is a much similar algorithm to Q-learning, the main difference is that it is an
difference is that it is an on-policy algorithm in which agent on-policy algorithm in which agent interacts with the environment and updates the
interacts with the environment and updates the policy based on policy based on actions taken. It means that the Q-value is updated by an action
actions taken. It means that the Q-value is updated by an action performed by the current policy instead of the greed policy that maximizes Q-value. In
performed by the current policy instead of the greed policy that this perspective, it is relevant when an action of one IoT device will greatly influence the
maximizes Q-value. In this perspective, it is relevant when an action condition of the environment.
of one IoT device will greatly influence the condition of the
environment.
(Deep Q Network) Deep Q network (DQN) is developed to solve the (Deep Q Network) Deep Q network (DQN) is developed to solve the exploration problem
exploration problem for unseen states. In the case of Q-learning, the for unseen states. In the case of Q-learning, the agent is not capable of estimating
agent is not capable of estimating value for unseen states. To handle value for unseen states. To handle this generality problem, DQN leverages neural
this generality problem, DQN leverages neural network technology. As network technology. As a variation of the classic Q-Learning algorithm, DQN utilizes a
a variation of the classic Q-Learning algorithm, DQN utilizes a deep deep convolutional neural net architecture for Q-function approximation. In real
convolutional neural net architecture for Q-function approximation. environments not all possible states and conditions are not able to be observed.
In real environments not all possible states and conditions are not Therefore, DQN is more relevant than Q-learning or SARSA in real applications such as
able to be observed. Therefore, DQN is more relevant than Q-learning IoT. Since DQN could be used within only discrete action space, it can be utilized for
or SARSA in real applications such as IoT. Since DQN could be used traffic routing in the IoT network.
within only discrete action space, it can be utilized for traffic
routing in the IoT network.
(Deep Deterministic Policy Gradient) DQN has solved generality and (Deep Deterministic Policy Gradient) DQN has solved generality and exploration
exploration problem of the unseen or rare states. Deep Deterministic problem of the unseen or rare states. Deep Deterministic Policy Gradient (DDPG) takes
Policy Gradient (DDPG) takes DQN into the continuous action domain. DQN into the continuous action domain. DDPG is a deterministic policy gradient based
DDPG is a deterministic policy gradient based actor-critic, model- actor-critic, model-free algorithm. The actor decides the best action for each state and
free algorithm. The actor decides the best action for each state and critic is used to evaluate the policy, the chosen action set. In IoT applications, DDPG
critic is used to evaluate the policy, the chosen action set. In IoT can be utilized for the tasks that require controlled in continuous action spaces, such
applications, DDPG can be utilized for the tasks that require as energy-efficient temperature control, computation offloading, network traffic
controlled in continuous action spaces, such as energy-efficient
temperature control, computation offloading, network traffic
scheduling, etc. scheduling, etc.
5.1.4. Neural Network based algorithms for IoT 5.1.4. Neural Network based algorithms for IoT
(Recurrent Neural Network) Recurrent Neural Network (RNN) is a (Recurrent Neural Network) Recurrent Neural Network (RNN) is a discriminative type of
discriminative type of supervised learning model that takes serial or supervised learning model that takes serial or time-series input data. RNN is specifically
time-series input data. RNN is specifically developed to address developed to address issue of time dependency of sequential time-series input data. It
issue of time dependency of sequential time-series input data. It processes sequences of data through internal memory, and it is useful in IoT
processes sequences of data through internal memory, and it is useful applications with time-dependent data, such as identifying time-dependent patterns of
in IoT applications with time-dependent data, such as identifying (Long Short Term Memory) As an extension of RNN, Long Short Term Memory (LSTM) is
time-dependent patterns of sensor data, estimating consumption a discriminative type of supervised learning model that is specialized for serial or time-
behavior over time, etc. series input data as well [Hochreiter]. The main difference of LSTM from RNN is that it
utilizes the concept of gates. It actively controls forget gates to prevent the long term
(Long Short Term Memory) As an extension of RNN, Long Short Term time dependency from waning. Therefore, compared to RNN, it is more suitable for data
Memory (LSTM) is a discriminative type of supervised learning model with long time relationship and IoT applications requiring analysis on the long lag of
that is specialized for serial or time-series input data as well dependency, such as activity recognition, disaster prediction, to name a few [Chung].
[Hochreiter]. The main difference of LSTM from RNN is that it
utilizes the concept of gates. It actively controls forget gates to
prevent the long term time dependency from waning. Therefore,
compared to RNN, it is more suitable for data with long time
relationship and IoT applications requiring analysis on the long lag
of dependency, such as activity recognition, disaster prediction, to
name a few [Chung].
(Convolutional Neural Network) Convolutional neural network (CNN) is (Convolutional Neural Network) Convolutional neural network (CNN) is a discriminative
a discriminative type of supervised learning model. It is developed type of supervised learning model. It is developed specifically for processing 2-
specifically for processing 2-dimensional image data by considering dimensional image data by considering local connectivity, but now generally used for
local connectivity, but now generally used for multidimensional data multidimensional data such as multi channel sound signals, IoT sensor values, etc. As
such as multi channel sound signals, IoT sensor values, etc. As in in CNN neurons are connected only to a small subset of the input and share weight
CNN neurons are connected only to a small subset of the input and parameters, CNN is much more sparse compared to fully connected network. However,
share weight parameters, CNN is much more sparse compared to fully it needs a large training dataset, especially for visual tasks. In CNN, a new activation
connected network. However, it needs a large training dataset, function for neural network, Rectified Linear Unit (ReLU), was proposed, which
especially for visual tasks. In CNN, a new activation function for accelerates training time without affecting the generalization of the network
neural network, Rectified Linear Unit (ReLU), was proposed, which [Krizhevsky]. In IoT domains, it is often used for detection tasks that require some
accelerates training time without affecting the generalization of the visual analysis.
network [Krizhevsky]. In IoT domains, it is often used for detection
tasks that require some visual analysis.
(Variational Autoencoder) Autoencoder (AE) is a generative type (Variational Autoencoder) Autoencoder (AE) is a generative type of unsupervised
of unsupervised learning model. AE is trained to generate output to learning model. AE is trained to generate output to reconstruct input data, thus it has
reconstruct input data, thus it has the same number of input and the same number of input and output units. It is suitable for feature extraction and
output units. It is suitable for feature extraction and dimensionality reduction. Because of its behavior to reconstructing the input data at the
dimensionality reduction. Because of its behavior to reconstructing output layer, it is often used for machinery fault diagnosis in IoT applications. The most
the input data at the output layer, it is often used for machinery popular type of AE, Variational Autoencoder (VAE) is a generative type of semi-
fault diagnosis in IoT applications. The most popular type of AE, supervised learning model. Its assumptions on the structure of the data are weak
Variational Autoencoder (VAE) is a generative type of semi-supervised enough for real applications and its training process through backpropagation is fast
learning model. Its assumptions on the structure of the data are weak [Doersch]. Therefore, VAE is suitable in IoT applications where data tends to be diverse
enough for real applications and its training process through and scarce.
backpropagation is fast [Doersch]. Therefore, VAE is suitable in IoT
applications where data tends to be diverse and scarce.
(Generative Adversarial Network) Generative Adversarial Network (GAN) (Generative Adversarial Network) Generative Adversarial Network (GAN) is a hybrid type
is a hybrid type of semi-supervised learning model which contain two of semi-supervised learning model which contain two neural networks, namely the
neural networks, namely the generative and discriminative networks generative and discriminative networks [Goodfellow]. The generator is trained to learn
[Goodfellow]. The generator is trained to learn the data distribution deceive the latter network, so-called the discriminator. Then, the discriminator learns to
from a training dataset in order to generate new data which can discriminate the generated data from the real data. In IoT applications, GAN can be
deceive the latter network, so-called the discriminator. Then, the used in situations when something needs to be generated from the available data, such
discriminator learns to discriminate the generated data from the real as localization, way-finding, and data type conversion.
data. In IoT applications, GAN can be used in situations when
something needs to be generated from the available data, such as
localization, way-finding, and data type conversion.
5.2. Technologies for lightweight and real-time intelligence 5.2. Technologies for lightweight and real-time intelligence
As the era of IoT has come, some sort of light-weight intelligence is As the era of IoT has come, some sort of light-weight intelligence is needed to support
needed to support smart objects. Prior to the era of IoT, most of the smart objects. Prior to the era of IoT, most of the works on learning did not consider
works on learning did not consider resource-constrained environments. resource-constrained environments. Especially, deep learning models require many
Especially, deep learning models require many resources such as resources such as processing power, memory, stable power source, etc. However, it
processing power, memory, stable power source, etc. However, it has has been recently shown that the parameters of the deep learning models contain
been recently shown that the parameters of the deep learning models redundant information, so that some parts of them can be delicately removed to reduce
contain redundant information, so that some parts of them can be complexity without much degradation of performance [Ba], [Denil]. In this section, the
delicately removed to reduce complexity without much degradation of technologies to achieve real-time and serverless learning in IoT environments are
performance [Ba], [Denil]. In this section, the technologies to
achieve real-time and serverless learning in IoT environments are
introduced. introduced.
(network compression) Network compression is a method to convert a (network compression) Network compression is a method to convert a dense network
dense network into a sparse one. With this technology the network can into a sparse one. With this technology the network can be reduced in its size and
be reduced in its size and complexity. By pruning irrelevant parts or complexity. By pruning irrelevant parts or sharing redundant parameters, the storage
sharing redundant parameters, the storage and computational and computational requirements can be decreased [Han]. After pruning, the
requirements can be decreased [Han]. After pruning, the performance performance of the network is examined and the pruning process is repeated until the
of the network is examined and the pruning process is repeated until performance reaches the minimum requirements for the specific applications and use
the performance reaches the minimum requirements for the specific cases. As many parameters are removed or shared, the memory required is reduced, as
applications and use cases. As many parameters are removed or shared, well as computational burden and energy. Especially as most energy in neural network
the memory required is reduced, as well as computational burden and is used to access memory, the consumed energy dramatically drops. Although its main
energy. Especially as most energy in neural network is used to access limitation is that there is not a general solution to compress all kinds of network, but it
memory, the consumed energy dramatically drops. Although its main
limitation is that there is not a general solution to compress all rather depends on the characteristics of each network. However, network compression
kinds of network, but it rather depends on the characteristics of is still the most widespread method to make deep learning technologies to be
each network. However, network compression is still the most
widespread method to make deep learning technologies to be
lightweight and IoT-friendly. lightweight and IoT-friendly.
(approximate computing) Approximate computing is an approach to (approximate computing) Approximate computing is an approach to support deep
support deep learning in smart devices [Venkataramani], [Moons]. It learning in smart devices [Venkataramani], [Moons]. It is based on the facts that the
is based on the facts that the results of deep learning do not need results of deep learning do not need to be exact in many IoT applications but still valid
to be exact in many IoT applications but still valid if the results deep learning, not only the execution time but also the energy consumption are
are in an acceptable range. By integrating approximate computing into reduced [Mohammadi]. Based on the optimal trade-off between accuracy and run-
deep learning, not only the execution time but also the energy time or energy consumption, the network can be adjustably approximated. The network
consumption are reduced [Mohammadi]. Based on the optimal trade-off approximate technology can be well-used in such situations when the response time is
between accuracy and run-time or energy consumption, the network can more important than sophisticatedly analyzed results. Although it is a technology to
be adjustably approximated. The network approximate technology can be facilitate real-time and lightweight intelligence, the process of training models and
well-used in such situations when the response time is more important converting it to approximate network require some amount of resource. Therefore, the
than sophisticatedly analyzed results. Although it is a technology to approximated model can be deployed on smart devices but the learning and
facilitate real-time and lightweight intelligence, the process of approximation processes still need to take places on resource rich platforms.
training models and converting it to approximate network require some
amount of resource. Therefore, the approximated model can be deployed
on smart devices but the learning and approximation processes still
need to take places on resource rich platforms.
6. IANA Considerations 6. Use cases of AI/ML into IoT service
Many IoT service environments are equipped with camera, door lock, temperature
sensor, fire detector, gas detector, alarm, and so on. Each sensor is deployed with
particular purposes of each own to provide a specific service. However, there is a
problem that the sensor utilization is not high enough due to the provision of the service
using only a single sensor rather than multiple sensors and their mutual relations.
Therefore, the quality of the service provided is not high as well. To enhance the sensor
utilization and the service quality, all signals from various sensors should be analyzed
comprehensively to make the right decision. This section describes the use cases for
introducing AI / ML techniques in actual IoT service, utilizing multiple sensors.
6.1. Use case 1: Surveillance and Security in Smart Home
To minimize users inconvenience and ensure their safety, Surveillance and safety IoT
applications provided within smart homes require fast notification with good level of
precision IoT service quality for abnormal conditions detection. To do this, both data
preprocessing techniques and AI/ML technologies for analysis of anomalies with high
accuracy will be required.
(Training Data Generation) For Surveillance and Security, the processed data is
necessary because there is little data for anomalies and the ground truth values are
hardly available. Therefore, first, the steps to detect and calibrate the abnormal data are
essential before the anomaly data should be generated using domain knowledge. First,
Constructing simulators about targeted smart home and generating events against
external intrusions and then collecting the anomaly data can be considered.
Furthermore, based on the data collected in the actual environment, anomaly data
generation can proceed by breaking the relationship between sensors considering
possible links between them within any intrusive environment.
(AI/ML Algorithm) Characteristics of the IoT environment for surveillance and safety are
that a massive amount of data is collected and real-time responses are required. For
the kNN algorithm, since the more data sets, the stronger against the noise and the
higher the accuracy. If the appropriate dataset is used, the fast response can be
expected. It makes suitable for the service environment to be considered. In addition,
considering real-time data forecasting and analysis via LSTM, it is believed that
improved accuracy for real-time anomalies detection can be expected.
6.2. Use case 2: Smart Health in Smart Home
(To be continued)
7. IANA Considerations
This document requests no action by IANA. This document requests no action by IANA.
7. Acknowledgements 8. Acknowledgements
8. Contributors 9. Contributors
9. Informative References
[Hochreiter] S. Hochreiter and J. Schmidhuber, "Long short-term 10. Informative References
memory," Neural Comput., vol. 9, no. 8, pp. 1735-1780, Nov. 1997. [Hochreiter] S. Hochreiter and J. Schmidhuber, "Long short-term memory," Neural
Comput., vol. 9, no. 8, pp. 1735-1780, Nov. 1997.
[Chung] J. Chung, C. Gulcehre, K. Cho, and Y. Bengio, "Empirical [Chung] J. Chung, C. Gulcehre, K. Cho, and Y. Bengio, "Empirical evaluation of gated
evaluation of gated recurrent neural networks on sequence modeling," recurrent neural networks on sequence modeling," arXiv preprint arXiv:1412.3555v1 [cs.NE],
arXiv preprint arXiv:1412.3555v1 [cs.NE], 2014. 2014.
[Krizhevsky] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet [Krizhevsky] A. Krizhevsky, I. Sutskever, and G. E. Hinton, "ImageNet classification with
classification with deep convolutional neural networks," in Proc. Adv. deep convolutional neural networks," in Proc. Adv. Neural Inf. Process. Syst., 2012, pp.
Neural Inf. Process. Syst., 2012, pp. 1097-1105. 1097-1105.
[Doersch] C. Doersch, "Tutorial on variational autoencoders," arXiv [Doersch] C. Doersch, "Tutorial on variational autoencoders," arXiv preprint
preprint arXiv:1606.05908v2 [stat.ML], 2016. arXiv:1606.05908v2 [stat.ML], 2016.
[Goodfellow]. I. Goodfellow et al., "Generative adversarial nets," in [Goodfellow]. I. Goodfellow et al., "Generative adversarial nets," in Proc. Adv. Neural Inf.
Proc. Adv. Neural Inf. Process. Syst., 2014, pp. 2672-2680. Process. Syst., 2014, pp. 2672-2680.
[Ba] J. Ba and R. Caruana, "Do deep nets really need to be deep?" in [Ba] J. Ba and R. Caruana, "Do deep nets really need to be deep?" in Proc. Adv. Neural Inf.
Proc. Adv. Neural Inf. Process. Syst., Montreal, QC, Canada, 2014, pp. Process. Syst., Montreal, QC, Canada, 2014, pp. 2654-2662.
2654-2662.
[Denil] M. Denil, B. Shakibi, L. Dinh, N. de Freitas, and M. Ranzato, [Denil] M. Denil, B. Shakibi, L. Dinh, N. de Freitas, and M. Ranzato, "Predicting parameters
"Predicting parameters in deep learning," in Proc. Adv. Neural Inf. in deep learning," in Proc. Adv. Neural Inf. Process. Syst., 2013, pp. 2148-2156.
Process. Syst., 2013, pp. 2148-2156.
[Han] S. Han, J. Pool, J. Tran, and W. Dally, "Learning both weights [Han] S. Han, J. Pool, J. Tran, and W. Dally, "Learning both weights and connections for
and connections for efficient neural network," in Proc. Adv. Neural efficient neural network," in Proc. Adv. Neural Inf. Process. Syst., Montreal, QC, Canada,
Inf. Process. Syst., Montreal, QC, Canada, 2015, pp. 1135-1143. 2015, pp. 1135-1143.
[Venkataramani] S. Venkataramani, A. Ranjan, K. Roy, and A. [Venkataramani] S. Venkataramani, A. Ranjan, K. Roy, and A. Raghunathan, "AxNN:
Raghunathan, "AxNN: Energy-efficient neuromorphic systems using Energy-efficient neuromorphic systems using approximate computing," in Proc. Int. Symp.
approximate computing," in Proc. Int. Symp. Low Power Electron. Low Power Electron. Design, ACM, 2014, pp. 27-32.
Design, ACM, 2014, pp. 27-32.
[Moons] B. Moons, B. De Brabandere, L. Van Gool, and M. Verhelst, [Moons] B. Moons, B. De Brabandere, L. Van Gool, and M. Verhelst, "Energy- efficient
"Energy- efficient ConvNets through approximate computing," in Proc. ConvNets through approximate computing," in Proc. IEEE Winter Conf. Appl. Comput. Vis.
IEEE Winter Conf. Appl. Comput. Vis. (WACV), Lake Placid, NY, USA: (WACV), Lake Placid, NY, USA: IEEE, 2016, pp. 1-8.
IEEE, 2016, pp. 1-8.
[Mohammadi] Mohammadi, Mehdi, et al. "Deep learning for IoT big data [Mohammadi] Mohammadi, Mehdi, et al. "Deep learning for IoT big data and streaming
and streaming analytics: A survey," IEEE Communications Surveys & analytics: A survey," IEEE Communications Surveys & Tutorials, 2018, pp. 2923-2960.
Tutorials, 2018, pp. 2923-2960.
[Kriegel] Kriegel, HansPeter, et al. "Densitybased clustering," [Kriegel] Kriegel, HansPeter, et al. "Densitybased clustering," Wiley Interdisciplinary
Wiley Interdisciplinary Reviews: Data Mining and Knowledge Discovery, Reviews: Data Mining and Knowledge Discovery, 2011, pp. 231-240.
2011, pp. 231-240.
Authors' Addresses Authors' Addresses
Jun Kyun Choi (editor) Jun Kyun Choi (editor)
Korea Advanced Institute of Science and Technology (KAIST) Korea Advanced Institute of Science and Technology (KAIST)
193 Munji Ro, Yuseong-gu, Daejeon 193 Munji Ro, Yuseong-gu, Daejeon
Korea Korea
Email: jkchoi59@kaist.ac.kr Email: jkchoi59@kaist.ac.kr
skipping to change at page 21, line 34 skipping to change at line 784
Korea Korea
Email: j89449@kaist.ac.kr Email: j89449@kaist.ac.kr
Min Kyung Kim Min Kyung Kim
Korea Advanced Institute of Science and Technology (KAIST) Korea Advanced Institute of Science and Technology (KAIST)
193 Munji Ro, Yuseong-gu, Daejeon 193 Munji Ro, Yuseong-gu, Daejeon
Korea Korea
Email: mkkim1778@kaist.ac.kr Email: mkkim1778@kaist.ac.kr
Gyu Myoung Lee
Liverpool John Moores University
Barkhill Rd, Merseyside, Liverpool L17 6BD
United Kingdom
Email: G.M.Lee@ljmu.ac.uk
 End of changes. 92 change blocks. 
629 lines changed or deleted 630 lines changed or added

This html diff was produced by rfcdiff 1.47. The latest version is available from http://tools.ietf.org/tools/rfcdiff/