1

Specification and Performance Analysis of

Wi-SUN FAN
Rei Hirakawa, Student Member, Keiichi Mizutani, and Hiroshi Harada, Member, IEEE

employs a physical layer compliant with IEEE 802.15.4-2015

Abstract—In recent years, extensive research has been conducted [4],[8], an international standard for low data rates and low-
on Internet of Things (IoT). Wireless Smart Ubiquitous Network power consumption. It primarily uses a frequency lower than 1
(Wi-SUN) has gained considerable attention as a wireless GHz band, which is called sub-1 GHz, and can transmit over a
communication standard for IoT. Wi-SUN Field Area Network
(Wi-SUN FAN) is a technical specification of Wi-SUN that can be
long distance of several hundred meters to one kilometer at a
implemented in both indoor and outdoor IoT communication power of approximately 20 mW [2] or higher. Furthermore, the
infrastructure with multi-hop routing. Although Wi-SUN FAN Carrier Sense Multiple Access with Collision Avoidance
version 1.0 (Wi-SUN FAN 1.0) has been standardized by IEEE (CSMA/CA) standardized in IEEE 802 15.4-2015 is adopted as
2857-2021, there have been no studies or reviews conducted on the the datalink layer protocol to reduce the radio interference
transmission performance of Wi-SUN FAN 1.0 regarding between the radio nodes within the same frequency [2]. Wi-
transmission success rate and delay time using computer
simulations and experimental evaluation environments involving
SUN has developed several technical specifications called
actual devices. In this study, the specifications of the Wi-SUN FAN profiles in Wi-SUN Alliance to flexibly support various
are reviewed, and the fundamental transmission performance, applications. The typical profiles include the Wi-SUN Home
such as average transmission success rate and average delay time, Area Network (HAN) and Wi-SUN Field Area Network (FAN)
is measured using computer simulation as reference data. An [7].
experimental evaluation environment involving actual devices is Wi-SUN HAN consists of two types of communication, as
developed to validate the characteristics evaluated by computer
simulation. The characteristics determined by the computer
shown in Fig. 1(a). The first is direct communication between
simulation and experimental evaluation environment are in good the electricity smart meter and Home Energy Management
agreement. Using the validated simulator, we evaluate the System (HEMS). The second is point-to-multipoint
transmission performance in the wireless IoT environment with communication between the HEMS and various devices, such
one border router and 100 routers randomly arranged in a flat as home appliances, solar panels, storage batteries, etc. In Japan,
square field with 4,000 m on a side. The average transmission the Wi-SUN HAN is widely adopted as a communication
success rate is approximately 1 at 1.00 × 10–1 s–1 or less.
Consequently, Wi-SUN FAN 1.0 can communicate with a higher
method between the electricity smart meter and HEMS, and
transmission success rate even when transmitting frequent IoT- more than 10 million meters with the function of Wi-SUN HAN
data, which is once every ten seconds. have already been introduced [9].
Wi-SUN Field Area Network (Wi-SUN FAN) is a
Index Terms—IEEE 2857, IEEE 802.15.4, multihop, RPL, Wi- communication specification used to transmit data acquired
SUN FAN. from outdoor-installed sensors, meters, and monitors to the
cloud and control these devices using the cloud based on the
analysis results of the collected data, as illustrated in Fig. 1(b).
I. INTRODUCTION In May 2016, Wi-SUN Alliance standardized the technical

T HE research and development (R&D), demonstration

testing, and commercialization of Internet of Things (IoT)
have been conducted extensively in recent years. IoT involves
communication specification as Wi-SUN FAN 1.0 [7], and the
Wi-SUN FAN 1.0 was also standardized as IEEE 2857 in June
2021 [10]. FSK is adopted as a transmission scheme for
wireless devices equipped with various "things", including physical layers with data rates of 50, 100, and 150 kbps in Wi-
sensors, meters, and monitors. The information collected from SUN FAN 1.0 by using frequency bands in the sub-1 GHz range.
the physical space is transmitted into a cyberspace comprising Furthermore, Wi-SUN FAN adopts frequency hopping (FH)
data management servers and databases, and new applications along with CSMA/CA to avoid interference in communication
are created by linking the collected data with other data [1],[2]. among different nodes. The network layer of Wi-SUN FAN
The Wireless Smart Ubiquitous Network (Wi-SUN) is a adopts IPv6 and supports multi-hop routing using the IPv6
wireless communication standard for IoT [2]–[7]. Wi-SUN Routing Protocol for Low-power and Lossy Networks (RPL)

This work was supported by the Ministry of Internal Affairs and The authors are with Department of Communications and Computer
Communications (MIC) in Japan under Grant Number JPJ000254 and the Engineering, Graduate School of Informatics, Kyoto University, Kyoto 606-
MIC/SCOPE under Grant Number JP196000002. (Corresponding author: 8501, Japan (e-mail: [email protected]).
Hiroshi Harada).
 2

Layer Protocols

Security IEEE 802.1X

Layer 4: Transport UDP / TCP
Layer 3: Network IPv6 / ICMPv6 / RPL / 6LoWPAN
LLC L2MESH (Option)
Layer 2: Data Link
MAC
IEEE 802.15.4-2015
Layer 1: Physical

Fig. 2. Protocol stack standardized as Wi-SUN FAN 1.0.

environment by actual devices was constructed to validate the

characteristics evaluated by computer simulation. Using the
(a) Wi-SUN HAN validated simulator, we evaluated the transmission
performances of large scale Wi-SUN FAN networks with one
border router (BR) and 100 routers arranged in a fixed
environment and randomly arranged in a flat square field with
4,000 m on a side. To the best of our knowledge, there are no
other studies of validated reference data. The results of this
study can be used as reference data to evaluate the actual Wi-
SUN network and to design communication systems based on
Wi-SUN FAN 1.0.
The remainder of this paper is organized as follows. Sections
II and III explain the specification and routing scheme of Wi-
SUN FAN 1.0, respectively. Section IV evaluates the
(b) Wi-SUN FAN transmission success rates and transmission delay times of
packets in multiple small-scale Wi-SUN FAN topologies via
Fig. 1. Wi-SUN application. computer simulations. Section V validates the results of the
computer simulation by experimental evaluation involving
routing protocol [11]. The maximum number of multi-hop actual devices. Section VI evaluates the transmission
stages exceeds 20 [10]. The multi-hop routing uses an performances of large scale Wi-SUN FAN networks with one
intermediate node as a router to ensure secure connectivity even BR and 100 routers and discusses the feasibility of high
in an area where direct communication is difficult and frequent data transmission in IoT environment. Finally, Section
eliminates the dead zone. VII concludes this paper.
Since the technical specifications were developed in May
2016, there has been extensive research conducted on Wi-SUN II. WI-SUN FAN 1.0
FAN 1.0. Generally, these studies present: a method to increase
the rate of construction of a multi-hop network [3], performance A. Protocol Stack
comparison of the average transmission success rate between Fig. 2 presents a protocol stack in Wi-SUN FAN 1.0, which
the MAC layer protocol in Wi-SUN FAN and the TSCH is standardized as IEEE 2857-2021 [10]. Wi-SUN FAN defines
protocol [4], method to construct a communication path in a the physical (PHY), data link, network, and transport layers
multi-hop network that is robust to variation in the corresponding to layers 1 through 4 in the OSI reference model.
communication environment [5], and a method to manage FH The data link layer is further divided into the upper Logical
to increase the throughput [6]. In February 2019, Wi-SUN Link Control (LLC) sublayer and the lower Medium Access
Alliance certified interoperable radio devices between Control (MAC) sublayer. Hereafter, the MAC sublayer is
manufacturers compliant with Wi-SUN FAN 1.0 [7]. A radio referred to as the MAC layer in this study. The network and
device developed by the author’s research group was one of the transport layers provide communications with User Datagram
world's first certified radio devices [7]. Because the Wi-SUN Protocol (UDP) or Transmission Control Protocol (TCP) based
FAN system has not been standardized for a long time, there is on IPv6. Moreover, Wi-SUN FAN supports various security
no concise document with detailed standard specifications, and technologies, such as the IEEE 802.1X, IEEE 802.11i, and
there was also no reference data available to evaluate the Extensible Authentication Protocol Transport Layer Security
transmission performance of the Wi-SUN FAN network. (EAP-TLS), to realize secure communication. A network called
In this study, the specifications of the Wi-SUN FAN 1.0 are the Personal Area Network (PAN) is constructed in Wi-SUN
reviewed, and the fundamental transmission performance, such FAN 1.0 by connecting radio devices called “nodes” which
as average transmission success rate and average delay time, is directly communicate with each other.
measured using different network topologies in computer
simulations as reference data. An experimental evaluation
 3

B. PHY layer #1 #2 #1
The PHY layer of Wi-SUN FAN conforms to the IEEE 802
15.4-2015 standard [2],[8]. IEEE 802.15.4-2015 defines three Node 1 Data frame ACK
Time
types of modulation schemes for Smart Utility Network (SUN):
Frequency Shift Keying (FSK), Orthogonal Frequency C1(t) CN 1 CN 5 CN 4 CN 2 CN 1 CN 3
Division Multiplexing (OFDM), and Offset Quadrature Phase- R1(t) CN 1 CN 5 CN 2 CN 1 CN 3
Shift Keying (O-QPSK) [2],[8]. In Wi-SUN FAN 1.0, FSK is
adopted as a transmission scheme for physical layers with data Node 2 Data frame ACK
Time
rates of 50, 100, and 150 kbps [10]. The operating frequency,
C2(t) CN 3 CN 2 CN 5 CN 1 CN 4
maximum transmission power, and transmission rate are
determined according to the national regulations [2],[10]. R2(t) CN 3 CN 2 CN 4
Furthermore, the data frame in the physical layer used in Wi-
SUN FAN 1.0, which is called the Physical Layer Packet Data #1 Listening on Cy(t) #2 Exchanging packet
Unit (PPDU), stores the MAC layer frame in a payload field Backoff counter Data generation
called the Physical Layer Convergence Protocol Data Unit CCA
(PSDU) [2]. Receiving packet Transmitting packet

CN x: Channel number x
C. MAC layer Cy(t): Assigned unicast listening schedule of node y
Ry(t): Actual listening channel of node y
The MAC layer defines functions related to detecting
connectable PAN and participating in PANs, as well as media Fig. 3. Packet transmission in unicast.
access control functions for multiple nodes to share and
communicate with media, such as frequency, time, etc. The 2) Frequency hopping (FH)
MAC layer of Wi-SUN FAN uses CSMA/CA standardized in The FH scheme is utilized to avoid interference between the
IEEE 802 15.4-2015 as the media access control method [8]. links by switching multiple operational frequency channels [6].
Additionally, FH is utilized to avoid the collision of frames by Each node performs a receiving operation corresponding to a
periodically switching the channels to be used. The CSMA/CA channel schedule which specifies the order of channels to be
and FH are described in detail below. used. Wi-SUN FAN employs two transmitting modes in
channel scheduling: unicast and broadcast, which use a unicast
1) CSMA/CA schedule and a broadcast schedule, respectively. The unicast
CSMA/CA is used to reduce the packet collision probability schedule is generated based on the MAC address of the node,
by performing clear channel assessment (CCA), which detects and it differs for each node. The unicast schedules specify the
the communication of other nodes via channel sensing after channels to wait for each time interval, which is called the
waiting for a random length backoff period before sending the unicast dwell interval (UDI). Furthermore, the unicast
packet. The length of the backoff period is determined by schedules are shared among the nodes, and each node can
multiplying the unit backoff time by a random number ranging understand the unicast schedules of the neighboring nodes.
from 0 to 2BE–1, where BE represents a constant called the Conversely, broadcast schedules are generated based on the
backoff exponent. A node in the backoff period can receive values unique to PAN, called broadcast schedule identifiers
packets, and the backoff operation is aborted if a packet is (BSI), which are advertised to all nodes in PAN and are
received. The backoff operation is resumed after the packet is common among all the nodes belonging to PAN. Broadcast is
received, and the node waits for the remaining backoff period. accomplished by using a common channel defined by the
Additionally, if the channel is available, the packet is broadcast schedule for all nodes. Broadcast scheduling defines
transmitted without detecting any carrier using the CCA; a time interval, called the Broadcast Dwell Interval (BDI), for
however, if the carrier is detected and the channel is unavailable, each Broadcast Interval (BI) during which a broadcast is to be
the BE is incremented by 1, up to the maximum backoff performed. During the BDI, each node waits for reception on
exponent. If the number of backoffs is less than or equal to the the channel specified in the broadcast schedule in preference to
maximum number of backoffs, the node waits again. The packet the unicast schedule. Conversely, during the periods other than
is retransmitted if the number of backoffs, NB, exceeds the the BDI, each node awaits reception by the unicast schedule on
maximum number of backoffs and if the number of packet its own channel. When the packet is transmitted to the node
retransmission is less than or equal to the maximum number of performing FH, the operations in the unicast and broadcast are
retransmissions. A binary exponential backoff method is used presented as follows.
to increase the range of the backoff period by a factor of two,
which reduces the probability of packet collisions in CSMA/CA, a) For unicast:
where an acknowledge (ACK) frame is received within a Fig. 3 presents an example of the unicast packet transmission.
certain period after a packet is transmitted. Following packet generation, the transmitting node performs a
backoff operation using CSMA/CA while awaiting reception on
the channel based on its channel schedule. Once the backoff
 4

operation is completed, based on the shared unicast schedule

information, CCA is performed on the channel in which the
destination node is awaiting reception. Transmission is
performed if the channel is available. However, transmission by
unicast is not initiated during the time when the broadcast
channel is valid in the broadcast schedule, and it waits until the
destination node starts receiving information based on the
unicast schedule. Subsequently, the node which initiated the
communication continues communicating on the same channel
without switching until the packet is transmitted and the
reception confirmation is generated by the ACK frame. The
node awaits reception on a channel based on its unicast schedule
once again when the communication is completed.

b) For broadcast:
A broadcast packet is transmitted only for the BDI period,
during which a broadcast channel is allocated for each BI. The
transmitting node transmits on the channel defined by the
broadcast schedule. The channel function determines the
Fig. 4. Operation of a node in Wi-SUN FAN.
unicast and broadcast schedules. Wi-SUN FAN employs the
directed hash channel function (DH1CF) [10] as a channel
function. In the DH1CF, the channel schedule is obtained by • PAN advertisement (PA) frames: A frame that enables the
using the current time, schedule identifier, and the number of nodes joined to a PAN to inform the nodes that do not
channels to be used. For a unicast and broadcast schedule, the participate in the PAN, of the minimum information
MAC address of the node and PAN-specified BSI are utilized required for PAN discovery and selection.
as the schedule identifier, respectively. • PAN advertisement solicit (PAS) frames: A frame that
enables the nodes that have not joined a PAN to request a
node that has joined the PAN, to transmit the PA frames
D. Network layer
more frequently.
Wi-SUN FAN uses IPv6 as the network-layer protocol. In • PAN configuration (PC) frame: A frame that enables a
addition to reporting the error information during packet node participating in the PAN to present the information
processing, Internet Control Message Protocol for IPv6 of the group key, channel schedule, current time, etc.
(ICMPv6) is used to provide various functions, such as address • PAN configuration solicit (PCS) frame: A frame that can
resolution, multicast, and duplicate IPv6 address detection. Wi- be used to request the nodes that have not been
SUN FAN supports multi-hop communication using IPv6 and participating in the PAN to transmit the PC frames more
can transmit packets in both directions based on the constructed frequently to the nodes that have already participated in
path. A routing protocol called RPL is adopted for low-power the network.
and lossy networks (LLNs) in the network layer [10],[11], and
each node autonomously selects a parent node to construct; it 2) Connection status
then changes and maintains the necessary paths for multi-hop Fig. 4 illustrates each connection status and operation until a
communication corresponding to changes in the propagation node completes the network layer routing and transmits the user
environment. Section III presents a detailed description of RPL. data. The operation is explained in detail below.

E. Basic operation of inter-node communication 1. Join State 1: PAN selection

In Wi-SUN FAN 1.0, a PAN is developed by connecting This is the initial state of the node, and in this state, a node
nodes that directly communicate with each other. This section does not yet contain the information on the neighboring nodes
describes the various control frames used for detecting PANs or the PAN. The node transmits PAS frames at intervals
and participating in a PAN. Fig. 4 depicts the operation controlled by the trickle timer [12] and requests the node
performed by a node in each connection state, and the belonging to the PAN to transmit the PA frames. When a PA
underlying process is described below. frame is received at the node, the information on the
neighboring node is added to the Extensive Authentication
1) PAN control frame Protocol over LAN (EAPoL) candidate list. The EAPoL
In Wi-SUN FAN, each node primarily uses the following candidates list is used to determine the PAN with which the
four control frames to detect and participate in PANs. node attempts to connect. Once a node determines a connecting
node from the EAPoL list within the selected PAN, the
connecting status proceeds to Join State 2.
 5

Root
2. Join State 2: Certification
In this state, the node executes the IEEE 802.1X/802.11i
security flow and performs authentication and key acquisition.
If authentication and key acquisition are successfully completed,
the node is set to the same PAN ID as the EAPoL destination,
and the connecting status proceeds to Join State 3. If
authentication and key acquisition fail, the connecting status (a) Directed Acyclic Graph (DAG) (b) Destination Oriented DAG (DODAG)
returns to Join State 1.
Fig. 5. Examples of DAG and DODAG.

3. Join State 3: PAN Settings the storing mode and the non-storing mode. In the storing mode,
The node sends the PCS frames at intervals controlled by the each node retains a routing table for a lower-rank DODAG, data
trickle timer and requests PC frames to the nodes belonging to routing is forwarded based on the routing table, and the
the PAN. If a PC frame can be received, the connection status downstream route is established. In the non-storing mode,
proceeds to Join State 4. However, if a PC frame cannot be nodes other than the root node do not retain the routing table,
received even after a certain number of PCS frames have been and the route node determines the downstream routes and
transmitted, the connecting status returns to Join State 1. In this presents this information to the other nodes. Wi-SUN FAN
case, the trickle timer is standardized by IETF for controlling utilizes RPL with the non-storing mode. Each node which
the transmission interval of the control frame of the network configures a multi-hop PAN in Wi-SUN FAN is categorized
[3],[12]; it is explained in more detail in Section III. B. into the following three types based on its function, as
illustrated in Fig. 1(b).
4. Join State 4: Route construction
In state 4, the node has completed participating in the PAN, • Border router (BR): The root node of a DODAG, which is
and network layer routing has been performed. If the node can located per PAN, is used to perform authentication, key
establish connectivity with the BRs within a certain period, the management, and access the wide area network (WAN).
connecting status proceeds to Join State 5, else the node returns Moreover, BR has a routing table for the entire DODAG to
to Join State 1 if connectivity cannot be established within a determine the downstream routes.
certain period. • Router: It can have a parent node and child nodes, and it
generates and forwards packets.
5. Join State 5: Connected
• Leaf: It has minimal functionality, such as packet-
In this stage, the network layer routing is complete. The PA
generation, transmission, and reception. Since it does not
and PC frames continue to be transmitted at intervals controlled
have the packet forwarding function, it becomes a terminal
by the trickle timer. If the node is disconnected with BR for a
node without a child node.
certain period, the status returns to Join State 1. Moreover, the
BRs consistently maintain Join State 5.
RPL assigns an ID called RPL instance-ID to each network,
where a set of DODAGs with the same RPL instance-ID is
III. ROUTING SCHEME BY RPL
called the RPL instance.
Wi-SUN FAN adopts RPL [5],[10],[11] as a routing protocol
in the network layer to support multi-hop communication using
IPv6. This section describes the network model in RPL and the B. Trickle timer
various control messages. Additionally, the routing metrics and The trickle timer [3], [12] controls the transmission interval
routing construction schemes defined in Wi-SUN FAN 1.0 are of the control frames of the network. It reduces the traffic
also described. generated by the control frames by exponentially increasing the
transmission interval when the network is stable. Conversely, if
the network is unstable, the trickle timer reduces the
A. Network model transmission interval to stabilize the network by frequently
RPL constructs a network of topologies based on a directed transmitting control frames.
acyclic graph (DAG); however, the Destination Oriented DAG
(DODAG) is directed without a cyclic graph. The DODAG C. Control message
comprises a root node without any outgoing edges. Fig. 5
In RPL, the following control messages are used for routing.
presents both the DAG and DODAG. In the RPL, the route from
any node in the DODAG to the root node is defined as the
• DODAG Information Object (DIO): Transmitted to obtain
upstream route, and the route from the root node to any other
the information required to select the parent and maintain
node in the DODAG is defined as the downstream route. Two DODAG. The nodes participating in the RPL network
principal methods are used to determine the downstream routes: broadcast DIO at the timing determined by trickle timer
 6

[3],[12]. The DIO contains information on the rank of the have been performed since the last ETX was calculated and
sender, and the node receiving the DIO selects the parents more than one minute has elapsed. Moreover, the ETX is
based on this information. The exception is that Leaf does calculated for a single transmission attempt immediately when
not send a DIO because it cannot have children. the node is initiated.
• DODAG Information Solicitation (DIS): Transmitted from In RPL, the Objective Function (OF) provides each node a
the nodes that do not participate in RPL network to request means to select and optimize the routes. In Wi-SUN FAN,
the DIO for nodes in the RPL network. These nodes minimum rank with hysteresis objective function (MRHOF) is
broadcast the DIS at regular intervals. used as an OF [5],[13]. In the RPL using MRHOF, the path-cost
• Destination Advertisement Object (DAO): Transmitted to and rank are adopted as indices to configure the route. These
construct and maintain a downstream route from the BR to terms are defined as follows:
the nodes that participate in the RPL network. In Wi-SUN
FAN, the non-storing mode RPL is used, and the nodes • Path cost: An indicator to evaluate the path to the BR. The
participating in the network present their parent information MRHOF preferentially selects a path with a lower path
to the BR by using DAO. The BR performs source routing cost. In Wi-SUN FAN, the path cost can be determined by
of the downstream routes based on the parent information the sum of the rank of the parent and the EWMA of ETX
of each node. The DAO is unicast at regular intervals from to the parent. The path cost when a node, N, selects the
each node to the BR. candidate parent, κ, is given as follows:
• DAO-ACK: Is used to verify whether the DAOs have 𝜗% (𝜅) = min{λ!"#$ (𝑁, 𝜅) + 𝜚(𝜅), 32768}, (3)
propagated to the BR. If the BR does not return the DAO- where λ!"#$ (𝑁, 𝜅) represents the EWMA of ETX for
ACK within a certain period, the node which transmitted the parent node, κ, at node, N. 𝜚(𝜅) represents the rank of the
DAO performs retransmissions. parent node, κ, which is notified by DIO.

D. Routing metric • Rank: An indicator of the relative position to the BR in the

RPL in Wi-SUN FAN utilizes the exponentially weighted network. The rank increases as a node moves away from
moving average (EWMA) of the received signal level (RSL) the BR. In Wi-SUN FAN, the rank of node, N, with the
and the expected transmission count (ETX) as link metrics parent of node, κ, is given as follows:
between the nodes. The EWMA is an exponentially weighted 𝜚(𝑁) = 𝐴B, (𝑁: BR)
moving average for a given series, X(t), and is given by the @𝜚(𝑁) = max [min {𝜚(𝜅) + 𝐴B , (𝑁: Other) (4)
following equation: 65535}, 𝜗% (𝜅)],
𝑋!"#$ (𝑡) = 𝑆𝑋(𝑡) + (1 − 𝑆)𝑋!"#$ (𝑡 − 1), (1) where 𝐴B represents the smallest rank, and 𝐴B = 128 in Wi-
where 𝑆 denotes a smoothing factor that satisfies 0 < 𝑆 < 1, SUN FAN.
and XEWMA(0) = X(0). In Wi-SUN FAN, 𝑆 = 1/8 [10]. RSL and
ETX are defined as follows: E. Routing procedures
• RSL: A metric based on the received power, with a bias of In RPL using MRHOF, routing with hysteresis properties is
174 added to it ranges from –174 to +80 dBm; this metric performed to preferentially select the paths with low path cost
is expressed as a value ranging from 0 to 254. RSLs are while preventing frequent changes in the network caused by
defined from the neighboring nodes to the self-node, and minor changes in the metrics. The route construction procedure
from the self-node to the neighboring nodes. The RSL for the upstream route is described below.
from a neighboring node to the self-node is obtained based
on the power received from the neighboring node. The 1. Creating a candidate parent Set: Each node creates a list
RSL from the self-node to the neighboring node is of neighboring nodes whose EWMA of RSL exceeds a
obtained based on the received power of the transmitted predetermined threshold. This list of neighboring nodes is
packet at the neighboring node and is transmitted from the called a candidate parent set. Additions and exclusions to
neighboring node to the self-node via an ACK frame. the candidate parent set are performed based on the
• ETX: A metric based on the transmission success rate of a conditional equation with the following hysteresis
frame, defined in Wi-SUN FAN as follows: characteristics:
λ = 128 ⋅ 𝑇⁄𝑆 = 128⁄𝜛 (𝑇/𝑆 < 8) • A node, N, adds a neighboring node, 𝜉, to the candidate
/ (2) parent set when both EWMAs of the bidirectional (i.e.,
λ = 1024, (𝑇/𝑆 ≥ 8)
downstream and upstream) RSLs for the neighboring
node, 𝜉, satisfy the following equation:
where T denotes the number of transmission attempts of the 𝜄(𝑁, 𝜉) > 𝑋W + 𝑌W − 𝑍B, (5)
frame, S denotes the number of reception times for ACK frame,
where 𝜄(𝑁, 𝜉) represents the EWMA of RSL, 𝑋W
and 𝜛 symbolizes the transmission success rate. The EWMAs
denotes the vendor-dependent minimum receiver
of RSL and ETX are calculated for all the packets
sensitivity level, 𝑌W denotes the constant which defines
acknowledged by the ACK frame. The ETX is calculated and
the EWMA of ETX is updated when four or more transmissions the threshold for adding the candidate parent set, and
 7

𝑍B denotes the hysteresis constant. In Wi-SUN FAN, 𝑌W

= 10 and 𝑍B = 3. Node #1 Node #N
• The node, N, excludes its neighboring node, 𝜉, from Application layer Application layer
the candidate parent set when both EWMAs of the Network layer Network layer
bidirectional RSLs for the neighboring node, 𝜉, satisfy MAC layer MAC layer
the following equation: Physical layer Physical layer
𝜄(𝑁, 𝜉) < 𝑋W + 𝑌W − 𝑍B. (6)

2. Calculating the path cost: The node derives the path cost Propagation path
𝜗% (𝜅) for the candidate parents in the candidate parent set
by using the rank 𝜚(𝜅) notified by the DIO. Information sharing register

3. Parent decision: Selects the candidate parent for which the Exchanging signals
path cost 𝜗% (𝜅) is the lowest in the candidate parent set. Exchanging packets and node information
Since the MRHOF has a hysteresis function, the node
maintains a connection with the current parent if the Fig. 6. Configuration of the computer simulation.
difference between the path costs for the current parent
and the new parent is less than the threshold 𝑇W. In MRHOF, a component which corresponds to a generated Wi-SUN FAN
the parent selection is performed in the following cases: node depending on the number of nodes to be simulated. Each
• The path cost for parent and candidate parents is node is independently operated and cannot directly refer to the
information of the other nodes. A node provides the physical,
updated.
MAC, network, and application layer functions, which are
• A new candidate parent is added
defined below.
A downstream route is established after establishing an Application layer: Generates UDP packets based on the packet
upstream route. The route development procedure for the generation rate. The generated packets are forwarded to the
downstream route is outlined below. network layer, and the received UDP packets are recorded.
1. The node transmits a neighbor solicitation (NS) and Network layer: Performs routing using RPL. The packets
registers its own IPv6 address with its parent, allowing the transmitted from the application layer are transmitted to the
BR to return the DAO-ACK. destination based on the route.
2. The node unicasts a DAO containing its IPv6 address
along with its parent' to the BR. MAC layer: Transmits packets using the CSMA/CA. FH is
3. The BR registers information in the DAO with the routing performed by generating various channel schedules are
table. Moreover, it determines the downstream route and generated. Additionally, the ETX and RSL used as link metrics
transmits the DAO-ACK back to the node. in the RPL are calculated and transmitted to the network layer.
4. The routing of the downstream route is completed when
the node receives the DAO-ACK. The DAO is Physical layer: Receives a transmission request from the MAC
retransmitted if DAO-ACK cannot be received within a layer and transmits the signal. In this layer, the delay time
certain period. associated with the reception switching of the radio is simulated.
Additionally, the propagation path is monitored, and the start of
IV. PERFORMANCE ANALYSIS IN MULTIPLE SMALL-SCALE the packet reception and collision are determined. The
WI-SUN FAN TOPOLOGIES BY COMPUTER SIMULATION determination of the collision is made as packet loss if CINR
This section evaluates the fundamental transmission (Carrier to interference plus noise power ratio) of the desired
characteristics in multiple small-scale Wi-SUN FAN networks packet is less than or equal to a predetermined threshold, and as
using RPL. In this study, a computer simulator was developed packet transmission success if the CINR is greater than the
for Wi-SUN FAN to evaluate different transmission threshold.
characteristics, such as the transmission success rate and delay
time for packet generation rate. The propagation path defines a radio wave propagation model
that attenuates signals based on the path model set beforehand,
A. Overview of computer simulator which corresponds to the positions of the nodes. The
Fig. 6 presents the configuration of the computer simulator information-sharing register registers the information that is to
used in this study. Originally, the computer simulator was be shared between the nodes. If a node requires information
developed for Wi-SUN FAN using C# based on [14] and can from another node, it consults the information sharing register.
flexibly set various parameters related to the system model, Examples of the information registered in the information
propagation model, and communication functions. The sharing register include information on the transmitted signals
computer simulation environment primarily comprises nodes, and packets, node location, and channel schedules.
propagation paths, and information sharing registers. A node is
 8

BR
B. Evaluation system
Router
A small-scale Wi-SUN FAN system is evaluated with star
and tree topologies using a total of 20 nodes, one BR, and 19
routers, to compare the characteristics with the actual system.
The evaluation assumed an upstream route where 19 routers
transmit their information to the BR. Fig. 7 presents all the
network topologies evaluated in this study. A propagation (a) Star topology
model which does not result in power loss between the
transmitting and receiving nodes is used in the computer BR
simulation. To avoid disrupting the topology, MAC filtering is Router
performed to discard packets when the source MAC address is
specified in the MAC layer. The MAC filters can be set
individually for each node. In a star topology network, the BR
is configured to receive packets from all nodes. Additionally,
the router is set to be able to only receive packets from the BR
by MAC filtering.
In a tree topology network, each node is configured to only
communicate with a particular node through MAC filters, and
a particular path is constructed. In total, 19 routers are divided (b) Tree topology
into two groups based on the routing of the packets transmitted
Fig. 7. Topology used in this study.
from the BR. A group where the number of transferring packets
from a node to BR is one is set as Group 1, and five routers were TABLE I
allocated. Subsequently, a group where the number of PARAMETERS RELATED TO THE NODE
transferring packets from a node to BR is two is set as Group 2, Parameter Value
Number of routers 19 (Section IV), 100 (Section VI)
and 14 routers were allocated. The BR can only receive packets
Transmit power 13 dBm (Section IV)
from routers in Group 1 by MAC filtering. A router in Group 1
Number of target data packets per node 100
can receive packets only from the BR, and the nodes in Group Data packet length 340 bytes
2 which are the child nodes of the router. A router in Group 2 Packet generation rate
–
From 1.00 ×10 2 to 1.00 s 1
–
can only receive packets from the node in Group 1 which is the Data rate 150 kbps
parent node of the router. RSSI threshold –104 dBm
In computer simulations, the transmission is evaluated based CCA threshold –84 dBm
on the assumption that the development of PAN is completed. Packet buffer size 15
Routing by RPL is initiated, and each node starts to generate
data packets independently. UDP is used to generate data TABLE II
PARAMETERS RELATED TO THE MAC LAYER
packets after a certain period; the data packets are generated at
Parameter Value
regular intervals based on the specified frequency. In this study, UDI 250 ms
such a data packet is simply referred to as a “packet”. The BI 1s
packets are forwarded to the BR using unicast according to the BDI 0.1 s
constructed upstream route. To evaluate the steady-state Number of channels 14, 1
characteristics of the network, the measurement was performed Minimum backoff exponent 4
after the 50th packet was generated at the node. The generation Maximum number of backoff 4
of the packet continues until the simulation is completed, and Maximum number of backoff 5
the simulation continues until there are no packets left to be Maximum number of retransmissions 4
measured from the packet buffer of all the nodes. Moreover, a Unit backoff time 5.3 ms
ACK frame length 72 bytes
node that fails to communicate with the BR and moves to Join
State 4 or less does not send a packet, and in such a case, the TABLE III
packet generated at the node is discarded. Tables I-III present PARAMETERS RELATED TO THE RPL
the parameters used in this study. The number of channels was Parameter Value
used for 1 and 14 to determine the effect of FH. DIS sending interval 30 s
Minimum DIO trickle timer Imin 1.024 s
Maximum DIO trickle timer Imax 7
C. Metrics for evaluation DAO sending interval 600 s
In this study, the transmission success rate and delay time are Maximum number of DAO retransmission 5
used as indices to evaluate the transmission efficiency in the DAO retransmission interval 10 s
Neighbor solicitation sending interval 600 s
network. The transmission success rate is defined as the ratio of
Size of candidate parent set 4
the number of packets successfully received in the BR to the
number of packets generated in the node. The delay time node until it is received at the BR. However, the packets that
includes the average time since a packet was generated at the have not reached the BR are excluded from the delay time
 9

1 100
UDP
DAO
DAO-ACK
0.8 80

Packet transmission rate (s–1)

Simulation (14ch) NS
Average success rate

Simulation (1ch) ACK

Experiment (14ch) Others
0.6 Experiment (1ch) 60

0.4 40

0.2 20

0 0
10–2 10–1 100 0 500 1000 1500 2000 2500 3000
Packet generation rate (s–1) Time (s)

Fig. 8. Average transmission success rate (star topology). Fig. 10. Breakdown of the number of packets transmitted per unit time when
the packet generation rate is 1.00 s‒1 and the number of channels used is one
(Star topology, simulation).
0.2 Simulation (14ch)
Simulation (1ch)
Experiment (14ch)
TABLE V
Experiment (1ch) BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE
0.15 PERIOD WHEN UDP PACKETS ARE TRANSMITTED (STAR TOPOLOGY,
Average delay (s)

SIMULATION, 1 CHANNEL)
Packet Packet count
Time
0.1
generation DAO
(s) UDP DAO NS ACK Other
rate (s–1) ACK
–2
1.00 ×10 26232.7 4982 820 820 1023 7434 4006
1.00 ×10–1 2623.9 4985 79 77 94 5212 402
0.05
3.33 ×10–1 787.7 5007 19 20 19 5048 120
5.00 ×10–1 420.0 4075 0 0 0 3990 65
1.00 209.7 4169 0 0 0 3984 40
0
10–2 10–1 100 range of packet generation rate was maintained at 1, regardless
Packet generation rate (s ) –1
of the presence or absence of FH. Fig. 9 presents the average
Fig. 9. Average delay time to successfully transmit packet (star topology).
delay time characteristics of all the nodes for the specified
packet generation rate. The average delay time is not lower than
TABLE IV approximately 65 ms even in a relatively low domain of packet
BREAKDOWN OF THE PACKET TRANSMISSION PROCESSING TIME
generation rate of 1.00 × 10–1 s–1 or less. This can be attributed
Items Minimum Average Maximum
to the delay due to the backoff time and the transmitting time of
Backoff 5.3 ms 42.4 ms 79.5 ms
CCA 0.128 ms the packets in the CSMA/CA.
Transmitter preparation 0.2 ms Table IV presents a breakdown of the packet transmission
Data packet transmission 18.133 ms processing time when there are no backoff retries or packet
Tack 1.1 ms retransmissions. The average packet transmission processing
ACK transmission 3.84 ms time is 65.801 ms, which is roughly consistent with the
Total 28.701 ms 65.801 ms 102.901 ms simulation evaluation results when packets are generated
infrequently. However, the average delay time increases when
calculation. Additionally, the number of child nodes and the the packet generation rate is 5.00 × 10–1 s–1 or more. The
number of contention nodes at each node are used as indices to average delay time is approximately 0.16 s when one channel is
evaluate the constructed routes. The number of child nodes is used and the packet generation rate is 1.00 s–1. This is caused
defined as the number of nodes connected to each node, and the by the backoff retries and packet retransmissions as the number
number of contention nodes is defined as the sum of the number of transmitted packets increases.
of nodes passing through one or more common routers in the Fig. 10 presents a breakdown of the number of packets
path to the BR. transmitted per unit time when one channel is used and the
packet generation rate is 1.00 s–1. In this study, the number of
packets transmitted per unit time is referred to as the packet
D. Results and Discussion
transmission rate. In Fig. 10, the horizontal axis represents the
First, the transmission characteristics in a star topology elapsed time from the start of the simulation, and the vertical
network are explained. Fig. 8 presents the average success rate axis represents the total packet transmission rates of all the
characteristics of all the nodes corresponding to the packet nodes by stacking various types of rates. In this figure, the
generation rate, where the packet generation rate is defined as packet transmission rate is measured every 30 seconds. Table V
the reciprocal of the number of packets generated per second. presents a breakdown of the number of packets transmitted in
In the computer simulation, the average success rate in the
 10

1 across the entire range of the evaluated packet generation rate

when the FH with 14 channels was used. Conversely, the
0.8 average transmission success rate drops in the region where
packet generation rate is greater than 7.00 × 10–1 s–1, and when
Average success rate

Simulation (14ch)
0.6
Simulation (1ch) packet generation rate is 1.00 s–1, the average transmission
Experiment (14ch)
Experiment (1ch) success rate is approximately 0.59, in the case when the number
0.4
of channels used is 1. This can be attributed to the packet
dropping due to buffer overflow at each node. Table VI presents
the average number of packets in buffers for five routers whose
0.2
parent node is the BR when the packet generation rate is 1.00 s–
1
. Here, the number of packets in the buffer was measured and
0 averaged every 5 seconds from the start of packet transmission.
10–2 10–1 100
The average number of packets in the buffer is one or less, and
Packet generation rate (s–1)
packets are transmitted without delay at each node when the
Fig. 11. Average transmission success rate (tree topology). number of channels to be used is 14. Meanwhile, when the
number of channels to be used is one, the average number of
TABLE VI
AVERAGE NUMBER OF PACKETS IN BUFFER OF EACH ROUTER packets in the buffer is 13 or more. Therefore, it is considered
Node index Packet count (14 channels) Packet count (1 channels) that the number of packets to be transmitted exceeds the actual
Router 1 0.6 13.9 number of packets to be transmitted, resulting in packet
Router 2 0.8 13.6 discarding due to buffer overflow. Besides, if the number of
Router 3 0.5 14.0 channels to be used is one, then simultaneous transmission by
Router 4 0.4 14.5
different nodes cannot be performed, and the number of packets
Router 5 0.2 13.5
transmitted is suppressed.
The average delay time characteristics of all nodes for packet
8 generation rate are illustrated in Fig. 12. The average delay time
7 is small when the packet generation rate is less than or equal to
1.00 × 10–1 s–1. The delay time is 126 ms when the packet
6
Simulation (14ch) generation rate is 1.00 × 10–2 s–1 and the number of FH channels
Average delay (s)

Simulation (1ch)
5 Experiment (14ch) is 14. In the tree topology network evaluated in this study, there
Experiment (1ch)
4 are five routers that perform packet transfer once and 14 routers
that perform packet transfer twice, and the average number of
3
transfers is approximately 1.74 times. Therefore, when packet
2 transmission is performed by the average packet transmission
1
processing time presented in Table IV at all the nodes, the
average delay time is approximately 114 ms. This value is
0
10–2 10–1 100
considered as a region where packets generate less frequently,
Packet generation rate (s–1)
and the delay time asymptotically approaches the same value.
The average delay time of all nodes increases with an increase
Fig. 12. Average delay time to successfully transmit packet (tree topology). in the packet generation rate. The average delay is
the period when UDP packets are transmitted. The total approximately 0.38 s if the packet generation rate is 1.00 s–1 and
transmission rate of the UDP packet of all the nodes is the FH with 14 channels is used. This increase in the average
approximately 19.9 s–1 (4169 packets/209.7 s) after 1860 delay time is attributed to the backoff retries or packet
seconds of initiating the data packet generation. Conversely, the retransmissions.
total packet generation rate of all the nodes is 19 s–1, which Fig. 13(a) presents a breakdown of the total packet
indicates that some packets have been retransmitted. In the star transmission rate for all the nodes when the packet generation
topology network, the transmission characteristics, such as the rate is 1.00 s–1 and the FH with 14 channels is used. Table VII
average transmission success rate and average delay time, do presents a breakdown of the number of packets transmitted in
not vary significantly corresponding to the number of channels the period when UDP packets are transmitted. It is evident that
used. In such a case, it is difficult to present interference the total of the UDP packet transmission rate at all nodes is
avoidance by FH even if multiple channels are available. approximately 38.9 s–1 (10210 packets/262.8 s) from the time
Second, the transmission characteristics in a tree topology period of 1860 s when the generation of the data packet was
network are discussed. Fig. 11 presents the average initiated until the measurement ends and the generation of UDP
transmission success rate characteristics of all the nodes packet ends. Conversely, it can be observed that the ACK frame
corresponding to the packet generation rate. The average transmission rate by the CSMA/CA in the same time interval is
transmission success rate was maintained at 0.999 or higher
 11

100 approximately 33.9 s–1. Since the UDP packet transmission rate
Packet transmission rate (s–1) is higher than the ACK frame transmission rate, a collision
UDP
80 DAO occurs in the UDP packet, and packet retransmission is assumed
DAO-ACK to occur. An increase in the delay time can be caused by a delay
NS
60 ACK
produced by a backoff retry or packet retransmission, as
Others explained earlier, a delay due to a wait time in the packet buffer,
40 or a packet transmission being suppressed owing to a limited
number of channels.
20
Fig. 13(b) presents a breakdown of the total packet
transmission rate for all the nodes when the packet generation
rate is 1.00 s–1 and the number of channels used is one. Table
0
0 500 1000 1500 2000 2500 3000 VIII presents a breakdown of the number of packets transmitted
Time (s)
in the period when the UDP packets are transmitted. It can be
(a) Number of packets sent per unit time (14 channels) observed that the total of the UDP packet transmission rate at
all the nodes is approximately 30.7 s–1 from the generation of
100
the data packet to the end of the measurement, and that the
Packet transmission rate (s–1)

UDP number of packet transmissions is suppressed compared to the

80 DAO case where the FH with 14 channels is used. There are conflicts
DAO-ACK
NS in the UDP packets and packet retransmission because the ACK
60 ACK frame transmission rate in the same interval is approximately
Others
25.4 s–1, which is smaller than the UDP packet transmission rate,
40 and the average delay time increases significantly to 7.20 s. It
can be observed from the above evaluation that in a tree
20 topology network, the interference between nodes can be
avoided by using an FH with multiple channels, and
0 transmission characteristics, such as average transmission
0 500 1000 1500 2000 2500 3000 success rate and average delay time, are significantly improved.
Time (s)

(b) Number of packets sent per unit time (1 channel) V. PERFORMANCE VALIDATION OF COMPUTER SIMULATION
RESULTS BY EXPERIMENTAL EVALUATION
Fig. 13. Breakdown of the number of packets transmitted per unit time when
packet generation rate is 1.00 s‒1 and the number of channels used is 14 (Tree In this section, the same configuration as the Wi-SUN FAN
topology, Simulation). network evaluated by computer simulation is constructed using
actual radio devices, and the results of computer simulation are
TABLE VII
BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE validated by experimental evaluation.
PERIOD WHEN UDP PACKETS ARE TRANSMITTED
(TREE TOPOLOGY, SIMULATION,14 CHANNELS)
Packet Packet count A. Evaluation environment
Time
generation DAO Fig. 14 depicts the configuration of Wi-SUN FAN used in
(s) UDP DAO NS ACK Other
rate (s–1) ACK
1.00 ×10 –2
26223.1 8689 1465 1466 1478 12931 3999 the experimental environment. Here, the Wi-SUN FAN
1.00 ×10–1 2622.8 8735 89 140 148 9064 400 mounted on USB board was used as a router to configure the
3.33 ×10–1 630.0 7110 19 38 45 7087 91 Wi-SUN FAN networking. The USB board consists of a Wi-
5.00 ×10–1 420.0 7300 0 0 10 7021 62
SUN FAN module BP35C5 by ROHM Co., Ltd., chip antenna,
1.00 262.8 10210 0 0 14 8909 38
and U.FL-type connector for an external antenna, and it can be
TABLE VIII operated as a Wi-SUN FAN node or used for capturing packets
BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE in the module firmware to operate Wi-SUN FAN 1.0 developed
PERIOD WHEN UDP PACKETS ARE TRANSMITTED
(TREE TOPOLOGY, SIMULATION, 1 CHANNEL) in the author’s laboratory and certified by the Wi-SUN alliance.
Packet Packet count A control command can be executed via a micro USB terminal
Time
generation
(s) UDP DAO
DAO
NS ACK Other on the board for various operations, such as achieving flexible
rate (s–1) ACK
control of the PHY and MAC parameters, updating firmware,
1.00 ×10 –2
26229.8 8655 1431 1428 1472 12931 4003
1.00 ×10–1 2623.0 8655 136 134 145 9061 400 log acquisition, etc. An evaluation board equipped with a
3.33 ×10–1 786.9 8705 35 34 46 8763 119 Raspberry Pi 3 Model B connected to 10 USB boards operates
5.00 ×10–1 420.1 7256 0 0 12 6941 63 as a single group. Moreover, the USB boards on the evaluation
1.00 262.8 8069 0 0 12 6667 34
board can be connected to each other through a U.FL-type
connector. Additional USB boards can be connected to each
other by connecting multiple evaluation boards with coaxial
 12

RF cable
was set at 60 s to increase the frequency of updating the routing
table in the BR during the construction phase of the network.
IoT gateway
Once all the routers complete routing through RPL and verify
that they are registered in the routing table of the BR, they start
Evaluation board generating the packets using UDP. The packet generation rates
Raspberry Pi 3 are 1.00 × 10–1 s–1, 3.33 × 10–1 s–1, 5.00 × 10–1 s–1, and 1.00 s–1.
The packets are generated at regular intervals based on the
USB board USB board
specified packet generation rate. During the measurement,
USB board USB board
packets were captured at each node. The transmitting power of
USB board USB board
each node was set to zero dBm. The other parameters are
USB board USB board identical to those presented in Section VI.
USB board USB board

Subsequent evaluation boards

C. Metrics for evaluation
The metrics for the experimental evaluation were essentially
Fig. 14. Configuration of the Wi-SUN FAN used in the experimental identical those defined in Section IV. Moreover, the received
environment. power from the routers in the BR and the network-configuration
times are analyzed.
TABLE IX
RECEIVED POWER FROM ROUTERS AT THE BR D. Results and Discussion
Received power Received power
Node index Node index Table IX presents the received power from the routers in the
(dBm) (dBm)
Router 1 –42 Router 11 –41 BR. The received power was calculated based on the EWMA
Router 2 –42 Router 12 –42 for the RSL of each router, which is stored in the BR after
Router 3 –41 Router 13 –42 measuring 14 channels in the star network. Since the nodes
Router 4 –42 Router 14 –47 were connected by the evaluation board in the measurement, the
Router 5 –41 Router 15 –43 received power of each node in the BR was distributed from –
Router 6 –41 Router 16 –41
47 dBm to the upper limit of the measurement range of –41
Router 7 –41 Router 17 –41
Router 8 –41 Router 18 –41 dBm. All the nodes can communicate with each other with
Router 9 –41 Router 19 –41 sufficient received power since the minimum received power of
Router 10 –41 Wi-SUN FAN module achieves the required Packet Error Rate
(PER) is –104 dBm [2].
cables. In this experimental evaluation, the IoT gateway, NSS- First, a star topology network is analyzed. Fig. 15 presents
OTDR-GW-V0002 by Nisshin Systems, was used as the BR. the transition of the number of routers connected to the BR
The IoT gateway is equipped with a BP35C5 and can perform when the FH with 14 channels is used. The vertical axis
the functions of the BR, achieve flexible control, change PHY represents the number of routers that joined the network and the
and MAC parameters, perform log acquisition, and other number of routers that have selected the BR as their parent
functions. Additionally, the IoT gateway can handle nodes. Joining the network was completed when the join state
neighboring node information of up to 1024 units, and a large- of the router was five. The number of routers was counted in
scale FAN network can be constructed and evaluated. The intervals of 1 s. Following a period of 22 s, nine routers selected
evaluation boards and IoT Gateway are connected to a control the BR as the parent node. This is because the BR transmitted
Windows PC, which executes various commands based on the the DIO and the nine routers that received it selected the BR as
measurement scenarios, collects logs, and synchronizes the the parent node. Once the parent node is selected, the router
time between the USB boards. transmits a DAO and completes the joining of the network
when communication with the BR is confirmed. The network
construction was completed within a relatively short period of
B. Specifications of experimental evaluation
115 s in the small-scale network to be measured.
The transmission performance of 20 nodes were evaluated
with a BR and 19 routers. The nodes were connected via an
evaluation board, and two evaluation boards were connected to
the IoT gateways configured as a BR using coaxial cables.
Two types of networks were evaluated: star topology and
tree topology. For each network, the MAC filters, similar to
those shown in Section VI, were set and the nodes that can
be connected to each node were restricted.
When starting the measurement, the BR was started after all
the routers were started. The DAO transmit interval in the router
 13

20 is 115,200 bps. Since there are other inputs/outputs of control

information, it is assumed to be one of the primary delay factors.
Number of routers

15 Conversely, if the packet generation rate is 1.00 s–1 and only one
channel is used, the average delay time is approximately 0.204
10 s. Fig. 16 presents a breakdown of the packet transmission rate
in this case. The packets, which are present prior to the
5
Network constructed beginning of the UDP packet transmission, are the packets
Selecting BR as preferred parent
related to PAN construction, such as PA, PC, and packets
0
0 20 40 60 80 100 120 related to security flow in IEEE 802.1X. Table X presents a
Time (s) breakdown of the number of packets transmitted in the period
Fig. 15. Transition of the number of Router connected to BR (star topology, when UDP packets are transmitted. The total data packet
Experiment).
generation ratio for all nodes is 19 s–1 after the UDP packet
100 transmission begins; however, the UDP packet transmission
UDP
DAO
rate is approximately 19.3 s–1 (3279 packets/171.0s), indicating
80 DAO-ACK that retransmissions occurred for some packets. The
Packet transmission rate (s )
–1

NS
ACK retransmission due to such packet collisions is a factor which
Others
60 increases the delay time.
Second, a tree topology network is analyzed. Fig. 17 presents
40 the transition of the number of routers when the FH with 14
channels was considered. For five routers configured to select
20 the BR as the parent node by MAC filtering, all selected the BR
as the parent node after 223 s. The period until all routers
0
completed their participation in the network is 307 s. In the tree
0 500 1000 1500 2000 2500 3000 topology network evaluated in this study, once the routers that
Time (s)
Fig. 16. Breakdown of the number of packets transmitted per unit time when perform packet transfer join the network, the routers that
packet generation rate is 1.00 s‒1 and the number of channels used is one (star perform packet transfer twice join the network. Accordingly,
topology, Experiment). the period until the network construction is completed becomes
longer compared to the star topology network. Fig. 11 presents
TABLE X
BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE the average transmission success rate of all the nodes for the
PERIOD WHEN UDP PACKETS ARE TRANSMITTED specified packet generation rate. The average transmission
(STAR TOPOLOGY, EXPERIMENT, 1 CHANNEL) success rate for both number of channels of 14 and 1, was 0.998
Packet Packet count
generation
Time
DAO
or higher in the range in which the packet generation rate is 5.00
(s) UDP DAO NS ACK Other × 10–1 s–1 or lesser. Conversely, the average transmission
rate (s–1) ACK
1.00 ×10–1 1679.1 3182 38 38 57 3310 329 success rate drops significantly when the packet generation rate
3.33 ×10–1 575.3 3637 19 19 19 3672 103 is 1.00 s–1. It is approximately 0.86 when the number of
5.00 ×10–1 395.3 3394 0 0 3 3237 68
1.00 171.0 3279 0 0 0 3134 29 channels used was 14 with FH. Fig. 18 presents a breakdown of
the packet transmission rate in this case. Table XI presents a
Fig. 8 presents the average transmission success rate of all breakdown of the number of packets transmitted in the period
nodes corresponding to the packet generation rate. The average when the UDP packets are transmitted. Since there are five
transmission success rate was maintained at one for both cases, routers which perform packet transfer once and 14 routers
i.e., for 14 channels and one channel. Fig. 9 presents the average which perform packet transfer twice, the expected total UDP
delay time characteristics of all the nodes for the packet packet transmission rate is approximately 33 s–1 for all nodes
generation rate. From both figures, the characteristics when no packet retransmission occurs. Conversely, the UDP
determined by the computer simulation and experimental packet transmission rate shown in Fig. 18 is approximately 60.1
evaluation environment were in good agreement. In the s–1 since collisions occur in many UDP packets and
experimental results, the average delay time is approximately retransmission is performed.
0.126 s for a packet generation rate of 1.00 × 10–1 s–1 when 14 The average transmission success rate is approximately 0.68
channels are used. This is longer than the average packet when one channel is used. Fig. 19 presents the breakdown of
transmission processing time of 65.801 ms given in Table IV. the packet transmission rate in this case. Table XII presents a
This delay is caused by the various data processing times in the breakdown of the number of packets transmitted in the period
actual devices. The primary delay factors in the actual device when the UDP packets are transmitted. The UDP packet
include the processing time by the input/output of commands transmission rate during the packet transmission interval is
and logs, etc. via the serial interface. It takes approximately 27.8 approximately 30.3 s–1, which is less than the case when 14
ms to input the 200 bytes user data and output it at the receiving channels are used. This is because the number of transmitted
end since the data rate of the serial interface of the USB board packets is restricted by the limited number of channels. It can
 14

20
100

Packet transmission rate (s–1)

Number of routers

15 Network constructed
Selecting BR as preferred parent 80 UDP
10 DAO
DAO-ACK
60 NS
5 ACK
Others
40
0
0 50 100 150 200 250 300 350
Time (s) 20
Fig. 17. Transition of the number of Router connected to BR (Tree topology,
experiment).
0
100 0 500 1000 1500 2000 2500 3000
Time (s)
Packet transmission rate (s–1)

80 UDP
Fig. 19. Breakdown of the number of packets transmitted per unit time when
DAO packet generation rate is 1.00 s‒1 (Tree topology, experiment, 1 channels).
DAO-ACK
60 NS TABLE XII
ACK BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE
Others PERIOD WHEN UDP PACKETS ARE TRANSMITTED
40
(TREE TOPOLOGY, EXPERIMENT, 1 CHANNEL)
Packet Packet count
Time
20 generation DAO
(s) UDP DAO NS ACK Other
rate (s–1) ACK
–1
1.00 ×10 1629.9 5330 68 66 57 5548 376
0 3.33 ×10–1 501.8 5553 26 22 19 5544 179
0 500 1000 1500 2000 2500 3000
Time (s) 5.00 ×10–1 372.3 5667 0 0 0 5540 157
Fig. 18. Breakdown of the number of packets transmitted per unit time when 1.00 224.8 6803 0 0 0 6096 95
packet generation rate is 1.00 s‒1 (Tree topology, experiment, 14 channels).
In terms of the breakdown of the number of total packets, the
TABLE XI
BREAKDOWN OF THE NUMBER OF TOTAL PACKETS TRANSMITTED IN THE characteristics determined by the computer simulation and
PERIOD WHEN UDP PACKETS ARE TRANSMITTED experimental evaluation environment were in good agreement.
(TREE TOPOLOGY, EXPERIMENT,14 CHANNELS)
Packet Packet count
generation
Time
DAO
VI. PERFORMANCE ANALYSIS IN MULTIPLE LARGE-SCALE
(s) UDP DAO NS ACK Other WI-SUN FAN TOPOLOGIES BY VALIDATED SIMULATOR
rate (s–1) ACK
–1
1.00 ×10 1820.9 6045 106 106 59 6356 415 In this section, using computer simulations validated in
3.33 ×10–1 569.6 7311 43 45 21 6427 204
5.00 ×10–1 408.0 8082 0 0 12 6886 171
Section V, we evaluate the transmission performance of a Wi-
1.00 208.4 12419 0 0 0 6330 83 SUN FAN 1.0 in multiple large-scale topologies with 100
multi-hop transmissions, as expected in an IoT application,
such as smart meters.
be observed that some UDP packets have not been received due A. Transmission performance in a fixed topology
to conflicts, etc., primarily because the ACK frame
With reference to Fig. 7, a tree topology was formed using
transmission rate is approximately 27.5 s–1, which is less than
30 nodes of Group 1 connected to BR and 70 nodes of Group 2
the UDP packet transmission rate. Fig. 12 depicts the average
transmitted to BR via the nodes of Group 1. In the topology, it
delay time characteristics of all the nodes for packet generation
was presumed that the node of Group 1 is always connected to
rate. The average delay time is approximately 0.175 s when 14
BR and the node of Group 2 can be connected to any node
channels are used and approximately 0.187 s when one channel
belonging to Group 1.
is used for a packet generation rate of 1.00 × 10–1 s–1. This delay
Fig. 20 presents the average success rate characteristics of all
time represents the actual packet transmission delay time along
the nodes corresponding to the packet generation rate. The
with various data processing times in the actual devices similar
average transmission success was maintained at 0.999 or higher
to the star topology network. Conversely, the average delay
in the range smaller than approximately 1.50 × 10–1 s–1.
time is approximately 0.806 s when 14 channels are in use and
Moreover, Fig. 21 presents the average delay time
approximately 3.535 s when one channel is in use, and in both
characteristics of all the nodes for the specified packet
cases, the packet generation rate is 1.00 s–1, which is a relatively
generation rate. The delay time was maintained at
high packet transmission rate (one packet per second). The
approximately zero in the range smaller than approximately
significant increase in latency for channel 1 can be attributed to
1.50 × 10–1. Based on these results, the information generated
the backoff retries, packet retransmissions, latency in packet
once every 10 seconds at all nodes can be transmitted to BR
buffers, and limited number of channels.
even though 100 units are connected.
 15

Fig. 20. Average transmission success rate (fixed large-scale tree topology, 10 Fig. 23. Average transmission success rate (a large-scale tree topology
0 router nodes). assuming for outdoor installation, 100 router nodes).

Fig. 21. Average delay time to successfully transmit packet (fixed large-scale
tree topology, 100 router nodes). Fig. 24. Average delay time to successfully transmit packet (a large-scale tree
topology assuming for outdoor installation, 100 router nodes).

Fig. 22. Node arrangement example used for the evaluation.

Fig. 25. Constructed route (fixed large-scale tree topology, 100 router nodes).

B. Transmission performance assuming for outdoor Moreover, with regards to route construction using RPL, a
installation router can connect to other routers without limitation, and there
The evaluation is conducted by assuming that the IoT device is no upper limit on the number of child nodes that can connect
with Wi-SUN FAN is installed outdoors. An example of node to each router. Simulation parameters for Wi-SUN FAN are
layout used for the evaluation is depicted in Fig. 22. One BR listed in Tables I, II and III. As a model of radio propagation
was placed in the center of a 4,000 m square flat field, and 100 model between nodes, a two-ray ground reflection model [16]
routers were randomly positioned. The height from a router to was adopted, similar to [15].
the ground is uniformly distributed from 1–10 m. The height of Fig. 23 presents the average success rate characteristics of all
BR was 3 m. These installation models were referred from [15]. nodes corresponding to the packet generation rate. Same as Fig.
 16

19, the average transmission success was maintained at 0.992 every ten minutes, to IoT applications, which transmit data once
or higher in the range smaller than approximately 1.0 × 10–1 s– every ten seconds. In summary, Wi-SUN FAN will drive next-
1
. Moreover, Fig. 24 presents the average delay time generation wireless IoT communication. Future research
characteristics of all the nodes corresponding to the packet objectives include experimental evaluation using hundreds of
generation rate. Same as Fig. 20, the delay time was maintained Wi-SUN FAN actual devices and practical demonstrations in
at approximately zero in the range smaller than approximately various applications.
1.0 × 10–1.
Fig. 25 illustrates an example of a finally constructed
network topology. It can be observed that even when 100 nodes REFERENCES
of router are randomly positioned in a flat square field with [1] U. Raza, P. Kulkarni, M. Sooriyabandara, “Low power wide area
4,000 m on a side, the constructed tree topology is almost the networks: an overview,” IEEE Communication Surveys & Tutorials, vol.
19, no. 2, Sept. 2017.
same type of ones evaluated in Sections IV and VI.A, most
[2] H. Harada, K. Mizutani, J. Fujiwara, K. Mochizuki, and K. Obata., “IEEE
number of hops is one same as Sections IV and VI.A and some 802.15.4g based Wi-SUN communication systems,” IEICE Trans.
of them is two. Consequently, the fundamental evaluation of Commun., vol. E100-B, no. 7, pp. 1032–1043, Jul. 2017.
this study can be used as the reference data for performance [3] T. Junjalearnvong, R. Okumura, K. Mizutani, and H. Harada,
evaluation of large-scale Wi-SUN FAN networks. “Performance evaluation of multi-hop network configuration for Wi-SUN
FAN systems,” in Proc. IEEE CCNC, Jan. 2019, pp.1–6.
[4] K. Mizutani, R. Okumura, K. Mizutani, and H. Harada, “Coexistence of
VII. CONCLUSION synchronous and asynchronous MAC protocols for wireless IoT systems
In this study, evaluation environments were developed using in sub-gigaherz band,” in Proc. WF-IoT, Jun. 2020, pp. 1–6.
computer simulations and actual experimental devices certified [5] D. Hotta, R. Okumura, K. Mizutani, and H. Harada, “Stabilization of
multi-hop routing construction in Wi-SUN FAN systems,” in Proc. IEEE
by Wi-SUN alliance on the Wi-SUN FAN 1.0 network CCNC 2020, Jan. 2020, pp.1–6.
standardized by IEEE 2857 as a wireless communication [6] R. Wayong, R. Okumura, K. Mizutani, and H. Harada, “A scheduling
standard for IoT applications. The average transmission success scheme for channel hopping in Wi-Sun FAN systems toward data
rate and average delay time were evaluated. Moreover, it was throughput enhancement,” in Proc. IEEE VTC 2020-Spring, May 2020,
pp.1–5.
confirmed that the characteristics determined by the computer
[7] Wi-SUN alliance, “Wi-SUN Alliance,” https://www.wi-sun.org
simulation and the experimental evaluation environment
[8] IEEE Computer Society, “IEEE Std 802.15.4TM-2015,” Dec. 2015.
conform with each other, and that the constructed computer
[9] T. Kiuchi and T. Okabe, “Sensor information transmission using a smart
simulation environment simulates the experimental meter communication network to monitor electric power facilities,”
environment using actual devices. Furthermore, it is CIGRE Colloquium Study Committee D2, Helsinki, Finland, Jun. 2019.
demonstrated that the reference data of the Wi-SUN FAN 1.0 [10] IEEE SA Board of Governors Corporate Advisory Group (CAG), “IEEE
system can be obtained using these two evaluation Std 2857TM-2021,” July 2021.
environments. In this study, a network is considered with one [11] T. Winter, P. Thubert, A. Brandt, J. Hui, R. Kelsey, P. Levis, K. Pister, R.
Struik, JP. Vasseur, R. Alexander, “RPL: IPv6 routing protocol for low-
BR and 19 routers; however, in the star topology network, the power and lossy networks,” IETF RFC 6550, Mar. 2012.
average transmission success rate is maintained at one, [12] P. Levis, T. Clausen, J. Hui, O. Gnawali, J. Ko, “The trickle algorithm,”
regardless of the communication frequency. Conversely, in the IETF RFC 6206, Mar. 2011.
tree topology, the average transmission success rate is [13] O. Gnawali and P. Levis, “The Minimum Rank with Hysteresis Objective
maintained at 0.998 or higher at 5.00 × 10–1 s–1 or less. However, Function,” IETF RFC 6719, Sep. 2012.
when the packet generation rate is 1.00 s–1, the average [14] H. Harada and R. Prasad, “Simulation and software radio for mobile
communications,” Artech House, 2002.
transmission success rate for both the computer simulation and
[15] R. Hirakawa, R. Okumura, K. Mizutani, and H. Harada, “A Novel routing
the actual machine reduced to approximately 0.59 and 0.68, Method with Load-Balancing in Wi-SUN FAN Network,” in Proc. WF-
respectively. This indicates that Wi-SUN FAN 1.0 may be able IoT 2021, June. 2021.
to communicate with a higher transmission success rate even [16] A. Goldsmith, “Wireless communications,” Cambridge University Press,
when transmitting frequent IoT-data, which is once every few Aug. 2005.
seconds. To confirm the feasibility of WI-SUN FAN in large-
scale environment, we evaluated the transmission performance
in the wireless IoT environment with one BR and 100 routers
that were randomly arranged in a flat square field with 4,000 m
on a side by using the validated simulator. The average
transmission success rate was maintained at 0.992 or higher in
the range less than approximately 1.0 × 10–1 s–1.
Consequently, Wi-SUN FAN 1.0 can communicate with a
higher transmission success rate even when transmitting
frequent IoT-data, which is once every ten seconds. The results
in this study confirmed that Wi-SUN FAN can be effectively
used for smart metering systems, which transmit data once
 17

alliance, and WhiteSpace alliance. He is a cofounder of Wi-

Rei Hirakawa received an M.I. degree of SUN alliance and has served as the chairman of the board from
Graduate School of Informatics, Kyoto 2012 to 2019. He is currently a vice chair of IEEE 2857, IEEE
University, Japan in 2022. He received a 802.15.4aa and Wi-SUN alliance. He moreover was the chair
B.E. degree in electric and electrical of the IEICE Technical Committee on Software Radio (TCSR)
engineering from Kyoto University in and the chair of Public Broadband Mobile Communication
2020. He researched low power and lossy Development Committee, ARIB. He is also involved in many
network protocols for wireless smart other activities related to telecommunications. He has authored
ubiquitous networks (Wi-SUN) in Kyoto the book entitled Simulation and Software Radio for Mobile
university. Communications (Artech House, 2002). He received the
achievement awards in 2006 and 2018 and fellow of IEICE in
2009, respectively and the achievement awards of ARIB in
2009 and 2018, respectively, on the topic of research and
Keiichi Mizutani is an associate professor development on software radio, cognitive radio, and Wi-SUN.
of Kyoto University. He received a B.E.
degree in electric, electrical and system
engineering from the Osaka Prefecture
University, Japan, in 2007, and an M.E.
and Ph.D. degree in electric and electrical
engineering from the Tokyo Institute of
Technology, Japan, in 2009 and 2012,
respectively. He was an invited researcher at Fraunhofer
Heinrich Hertz Institute, Germany, in 2010. From April 2012 to
Sept. 2014, he was a researcher at National Institute of
Information and Communications Technology (NICT). He
currently researches the topics of physical layer technologies in
White Space Communications, Dynamic Spectrum Access,
Wireless Smart Utility Networks (Wi-SUN), and 4G/5G/6G
systems including OFDM, OFDMA, MIMO, and multi-hop
relay network systems. Since joining in NICT, he has been
involved in IEEE 802 standardization activities, namely
802.11af, 802.15.4m and 802.22b. He received the Special
Technical Awards from IEICE SR technical committee in 2009
and 2017, the Best Paper Award from IEICE SR technical
committee in 2010 and 2020, the Young Researcher's Award
from IEICE SRW technical committee in 2016, the Best Paper
Award from WPMC2017 and WPMC2020, and the Best Paper
Presentation Award (1st Place) from IEEE WF-IoT 2020.

Hiroshi Harada is a professor of the

Graduate School of Informatics, Kyoto
University, and an Executive Research
Director of Wireless Networks Research
Center, National Institute of Information
and Communications Technology (NICT).
He joined the Communications Research
Laboratory, Ministry of Posts and
Communications, in 1995 (currently,
NICT). Since 1995, he has researched software radio, cognitive
radio, dynamic spectrum access network, wireless smart
ubiquitous network (Wi-SUN), and broadband wireless access
systems. He also has joined many standardization committees
and forums in the United States as well as in Japan and fulfilled
important roles for them, especially IEEE 1900 and IEEE 802.
He was the chair of IEEE DySpan Standards Committee and a
vice chair of IEEE 802.15.4g, IEEE 802.15.4m, IEEE 1900.4,
and TIA TR-51. He was a board of directors of IEEE
communication society standards board, SDR forum, DSA