Content uploaded by Timo Nihtilä
Author content
All content in this area was uploaded by Timo Nihtilä on Dec 13, 2022
Content may be subject to copyright.
Energy Saving Router Rotation
Protocol for DECT-2020 NR
Timo Nihtilä
Magister Solutions Ltd
Jyväskylä, Finland
timo.nihtila@magister.fi
Tarmo Taipale
Nordic Semiconductor Ltd
Turku, Finland
tarmo.taipale@nordicsemi.no
Abstract—ETSI DECT-2020 New Radio (NR) is a new flexible
radio interface targeted to support a broad range of wireless
Internet of Things (IoT) applications. It has been shown to fulfill
both massive machine-type communications (mMTC) and
ultra-reliable low latency communications (URLLC)
requirements for 5th generation (5G) networks. DECT-2020 NR
is an autonomous wireless mesh network protocol where the
devices can choose to become routers, forwarding the data of
other devices in addition to their own data. Thus, a wireless mesh
network does not need separate base stations or a core network
architecture to extend coverage. This makes the deployment of
DECT-2020 NR networks easy, but with the cost of increased
energy consumption in the router nodes. Notably, the same
energy consumption is not inflicted upon the non-routing leaf
nodes whose operation is on the contrary very energy efficient.
This role-induced disparity in node energy consumption results
in the network not using the energy of its devices with maximal
efficiency. A method to alleviate this problem could be a network
rotation in which the roles of network nodes are periodically
rotated and the energy consumption is thus distributed more
evenly among the nodes. In this paper we propose a role rotation
protocol for DECT-2020 NR and evaluate its impact to the
expected lifetime and the performance of the network by system
level simulations.
Keywords—DECT-2020, NR, 5G, energy saving
I. INTRODUCTION
The DECT-2020 NR standard was released by the
European Telecommunications Standards Institute (ETSI) in
2020 [1]. It is intended for industrial Internet of Things (IoT)
applications requiring either massive machine-type (mMTC)
and/or ultra-reliable low latency (URLLC) communications.
DECT-2020 NR introduces a device-based cluster-tree
wireless mesh network (WMN), capable of forming an
autonomous network without the need for separate base
stations or core network. Nodes are capable of acting as
routers between their neighbors, forwarding data between
them in case direct connection is not possible.
WMN coverage can theoretically reach extreme distances
with minimal costs and effortless deployment. However,
forwarding data through the network with a reasonable delay
requires that each node passing the data onwards has an
acceptable duty cycle, meaning the node is able to receive and
transmit reasonably often so that the delay requirements are
met and the data does not accumulate to relaying nodes.
However, maintaining an active duty cycle consumes energy
resources which may be scarce if the network consists of
mainly battery-operated nodes, as is common in the IoT
domain.
Energy consumption is a well-known and well-studied
problem of WMNs. Comprehensive surveys of different
energy-efficient routing protocols proposed for wireless
sensor/mesh networks have been presented in literature e.g.
[2], [3]. The advantages and disadvantages of each considered
algorithm has been quite extensively analyzed. However, no
earlier work on the suitability or the performance of WMN
energy saving protocols with DECT-2020 NR yet exists.
The performance of DECT-2020 NR has been evaluated
earlier e.g. in [4]-[7] but none of the studies gave
consideration to the network energy consumption. In [8] it was
demonstrated that especially when serving delay sensitive
data, the lifetime of a battery-operated DECT-2020 NR
network is limited by the energy consumption of the router
nodes, due to their active duty cycle required to maintain low
latencies. On the other hand, the leaf nodes of DECT-2020 NR
were shown to be very energy efficient as they only need to be
awake intermittently. Due to this disparity in energy
efficiency, it is evident that if the DECT-2020 NR network
could distribute the energy consumption more evenly between
the nodes, the network lifetime would increase.
More even energy consumption distribution can be
achieved via node role rotation. The idea has been proposed
for WMNs earlier in [9] where the authors proposed a
Low-Energy Adaptive Clustering Hierarchy (LEACH)
protocol. LEACH utilizes a randomized rotation of
cluster-heads, which are elected head nodes of local sensor
node clusters. Cluster-heads collect the data from the nodes in
the cluster, aggregate it and transmit it to the target base
station, so their function is roughly similar to DECT-2020 NR
routers. A cluster-head node consumes more energy than a
regular node so in LEACH the cluster nodes periodically elect
a different node to act as a cluster-head to evenly distribute the
energy load.
Similar role rotation idea was also used in [10], where the
Span algorithm was introduced for 802.11 ad hoc networks. In
Span, each network node periodically makes an independent
decision on whether to sleep or stay awake as a coordinator
node which is a similar node as DECT-2020 NR router node in
a sense that it participates in forwarding the packets through
the network.
Based on the findings of these earlier WMN studies one can
expect network lifetime benefits from rotation also with
DECT-2020 NR standard but exactly how much benefit is
achievable cannot be concluded without implementing a
rotation protocol that can be fitted into the standard and
evaluating its performance using realistic modeling.
Compared to conventional WMN technologies, the energy
efficiency of DECT-2020 NR is improved by the use of
state-of-the-art physical layer techniques, such as Cyclic
Prefix Orthogonal Frequency Division Multiplexing
(CP-OFDM), turbo coding and Hybrid ARQ (HARQ) and
periodic Random Access Channel (RACH) duty-cycling.
These features make the comparison with earlier, more simpler
WMN technologies difficult and justify the need for a new
study.
In this paper we adapt the WMN role rotation functionality
to DECT-2020 NR and propose a simple protocol to rotate the
roles of network nodes without the need of specific election
processes or central coordination but by merely reusing the
mesh network organization process inherent in the standard.
Finally, we evaluate the actual effect of rotation to
DECT-2020 NR network lifetime and performance by
conducting dynamic system level simulations in a typical IoT
use case with mesh networks of different sizes. The results of
the study can be used to improve the energy efficiency of
DECT-2020 NR standard in future releases.
II. DECT-2020 NR MESH NETWORK ORGANIZATION
The DECT-2020 NR network organization process is
explained in the standard [1]. For a single network, there is
always one sink, i.e. a gateway node to the Internet. There can
be multiple sinks in a group of nodes but from the system
point of view the sinks form their own networks (cluster
trees), which co-exist but operate independently from each
other. In this study we concentrate on the operation of a single
network entity under one sink.
A diagram of the initial organization process in a simple
network is presented in Fig. 1. The process starts by the sink
selecting FT (Fixed Termination point) mode. In FT mode, a
node first performs background (BG) scanning, searching for
the least busy channel for operation. After finding a channel,
the FT starts cluster beacon transmissions on it.
All other nodes are in PT (Portable Termination point)
mode, where they initially scan channels for cluster beacons
and after detecting one or more of them, select from them the
sender FT for association. A PT having multiple FT
candidates makes the choice based on the route cost, a
parameter of the route info information element (IE) of the
cluster beacon message. The specification mandates that each
hop should increase the route cost by at least 1 but otherwise
the exact route cost calculation is left to implementation so
essentially the node is free to use any information available.
Selecting a proper route cost metric is essential for ensuring
the occurrence of role rotation in the network. We will discuss
this more in section III.
Fig. 1 A diagram of a simple network initial organization process of where tier
1 (T1) node A becomes a router (R) and T1 node B and tier 2 (T2) node C
become leaf nodes (L).
After a PT has successfully associated itself with an FT, the
PT switches to FT mode and performs FT mode operations,
i.e. BG scanning and beaconing, in order to advertise itself as
a potential parent to its nearby nodes and by that extend the
coverage area of the network to a new tier of nodes. The nodes
associated directly to the sink form tier 1, the nodes associated
with tier 1 nodes form tier 2 nodes, etc.
As this organization cycle continues from tier to tier, the
mesh network size can theoretically increase to extreme
distances. The cycle continues until all nodes within a certain
coverage area are associated with another node. In the
simulations we let all the nodes perform BG scanning after
their association and after all nodes have performed it, we
consider the initial organization phase finished and the
network complete, after which all nodes initiate traffic
transmissions.
Some nodes in the network are left as pure PT nodes, i.e.
leaf nodes. These nodes cancel their advertisement beaconing
after organization is finished and go to sleep, only to wake up
listening to their parent’s beacon or to transmit data.
III. NETWORK ROTATION PROCESS
In this section we explain how the DECT-2020 NR network
rotation is executed in this study. It should be noted that the
purpose of this paper is to describe the basic principles of the
rotation process and focus on the actual impact of it to the
network performance. Thus, we address the implications and
the mere feasibility of implementing the process in a real
network only superficially in this paper.
A diagram of a rotation process in a simple network is
depicted in Fig. 2. The protocol consists of the following
phases: association release, advertisement and reassociation,
which are explained in the following sections.
Fig. 2 A diagram of a rotation process in a simple network, where tier 1 (T1)
node A switches its role from router (R) to leaf (L) and node B vice versa.
A. Association release phase
Rotation is always triggered by the sink, which starts an
association release phase. The sink releases the associations to
its children by transmitting an Association Release message to
each of them. The release process then continues recursively
so that a node receiving an Association Release message from
its parent will forward the message also to its own children, if
such exist. Hence, the process continues down on all the
branches of the mesh network recursively until all associations
in the network are released and all nodes are left orphans.
It is to be noted that during the association release phase,
traffic generation still persists in every node, so the association
release causes an inevitable service interruption. Thus, the
periodicity of triggering the rotation should be selected so that
the interruption time is minimized whilst making sure that as
many nodes as possible get a chance to act as a router before
the first ones deplete their batteries. However, any reasonable
IoT use case assuming battery-operated nodes needs to have a
relatively low traffic intensity in order to ensure an adequate
node lifetime [8]. Low traffic intensity means a packet interval
of e.g. several hours per node and results in years of expected
router battery life while the service interruption due to rotation
is likely to be over in minutes, depending on the beacon
interval and the depth of the cluster tree network. Thus the
rotation occurring e.g. once per day is likely short enough to
allow the network to go through all possible router
constellations many times during the network lifetime. A
rotation interval sparse as this most likely has a negligible
effect on the service quality.
B. Advertisement phase
Upon becoming an orphan after the association release
phase, a node starts beacon scanning immediately to find a
new parent. In order to ensure that the roles of the network
nodes are truly rotated, all nodes who associate need to send
an advertisement in order to establish as many reassociation
opportunities as possible for the remaining orphans.
Upon associating with an FT, a PT node becomes an
advertising FT (AFT) where it temporarily sends beacons and
enables its RACH listening. Nearby orphans in beacon
scanning mode thus have an opportunity to hear the AFT and
associate with it. If AFT receives association requests during
its advertisement phase, it becomes a router and continues
beaconing. Otherwise it switches back to PT mode after the
advertisement phase and continues as a leaf node until the next
rotation. The duration of the advertisement phase is a network
parameter. In our simulations, we assumed one beacon period
duration for this. A longer advertisement phase would increase
the success rate of associations with AFT, but also its energy
consumption.
Contrary to operation during the initial organization phase,
AFTs do not perform BG scanning before starting the
advertisement beaconing. It is assumed that the channel
selection made during the initial organization phase remains
valid as almost all nodes transmitted beacons during it and
were therefore able to be heard and avoided by their
neighbors. A small risk of mutual interference between
neighbors after reassociation is always left, but in this study
we calculated the benefit of energy saving by not performing a
new BG scan after each reassociation greater than the risk of
potential interference.
C. Reassociation phase
The phase with the most impact on the rotation process is
the reassociation phase. Here the PTs select new parents after
they have received one or multiple cluster beacons.
An intuitive value to base the FT selection on would be the
received beacon signal strength. If PTs prioritize the candidate
FTs who transmitted the strongest beacons, it would
presumably minimize the error probability and the required Tx
power of transmissions between them. This would obviously
impact positively on the battery life in that sense. However,
assuming the positions of the nodes in the network do not
change, basing the selection on beacon signal strength would
mean that the nodes would always select the same routers as
parents, whenever there was an attempt to reorganize the
network. This would cement the roles in the network and
effectively eliminate the possibility of rotation.
Generally, in order to achieve rotation of roles in the
network, the FT selection should be made based on a metric
which is not fixed throughout the lifetime of the network. In
order to achieve as even energy consumption among the nodes
as possible, we propose that the FT selection should prioritize
associations to FTs with the highest battery level. This way
after the rotation, nodes with low battery level would be more
likely to end up in leaf role, in which they can save energy and
stay connected to the network longer.
Router battery state is not specifically embedded into the
cluster beacon message in current DECT-2020 NR
specifications. However, implementations using the current
specifications could base their reported route cost metric value
to the inverse of their battery level and this would have the
same effect. Another option would be to introduce a new
battery level IE to the beacon message which then could be
utilized by the PT nodes in FT selection along with the route
cost metric.
IV. SIMULATIONS
For this study we used an open-source Network Simulator 3
platform [11] into which we implemented a module for
DECT-2020 NR. The module consists of the MAC and PHY
layers implemented according to ETSI TS 103 636 series
specifications [1]. The details of the simulator modeling are
explained in [8]. We added the periodic rotation process
explained in section III and modified the FT selection process
of the simulator accordingly.
TABLE I. SIMULATION PARAMETERS
Parameter
Value
Scenario radius
800 m - 5 km
Frequency
1.89 GHz
DECT channels
10
Propagation model
ITU Urban Macro
Shadowing
Disabled
Channel model
5G spatial channel model [13]
Number of nodes
[1000, 3000, 5000]
Number of sinks
1
No of simulation
drops per case
20
Simulation duration
24 h
Antenna model
1x1 omni, 0 dBi gain
Noise figure
7 dB
Beacon Tx power
23 dBm
Minimum beacon quality
for association
3 dB
Route cost
Inverse of FT battery level
Power control
Pathloss based: TxP [dBm] =
min (23, -68 - 0.7 * path gain)
Max HARQ retransmissions
9
HARQ processes per node
2
SCS scaling factor
1
FFT scaling factor
1
LBT min CW
8
LBT max CW
64
Table I shows the general simulation parameters and
assumptions. We simulated a DECT-2020 NR network in an
IoT scenario consisting of different amounts of nodes, from
1000 to 5000, to cover different network loads. Nodes were
randomly positioned in each simulation drop. The centermost
node was selected as the sink in each drop. Scenario radius
was also varied in order to capture the effects of different sized
mesh topologies with different amounts of routers and tiers.
As found in [8], a battery-operated DECT-2020 NR mesh
network is reasonable only with low traffic intensity and
relatively loose delay requirements. A high traffic intensity
severely decreases the network lifetime (resulting in mere
days of operation time), to which no energy saving mechanism
is able to do much about. Thus, in the simulations, we
assumed a general battery-operated IoT sensor network use
case with relatively low traffic intensity and set the parameters
to minimize node energy consumption whilst still ensuring
adequate service. The details of the selected use case and its
traffic parameterization are shown in Table II.
In the result analysis, we focused on the impact of router
rotation to the node battery life expectation and to system
performance in terms of end-to-end packet delay and the
service rate, which is the percentage of the transmitted data
that is successfully received by the target node (the sink).
End-to-end delay was measured from the time a packet was
generated in the source node to the time when it was
successfully received at the target node.
Each node transmits a packet to the sink once per 24 hours.
The rotation period was set to 2 hours, a short period in
relation to expected node lifetime, to ensure some rotations
occur during the 24 h simulation time. During the rotation
period, the network as a whole relays on average 83-417
packets through (depending on the number of nodes) and, with
a beacon interval of 32 seconds, each router transmits 225
beacons and listens to as many RACH occasions.
TABLE II. SIMULATION USE CASE
Parameter
Value
Real-life use case
examples
Smart energy demand response
management (DRM), Smart transport
road condition monitoring, Smart
agriculture irrigation / fertilization /
pest control [12]
Tolerable delay
Few minutes
Beacon interval
32 s
Router RACH
period
1 x 20 ms per beacon period
Traffic
100 B data packet / 24 h / node
Rotation
triggering period
2 hours
Traffic direction
Unidirectional, targeting sink
Table III shows the used energy parameters. Expected node
battery life for each node was calculated from their
experienced energy consumption during the simulation, which
was defined by the electric current usage of the radio states
they were in. State transitions did not consume any additional
energy. The used currents in different radio states were based
on actual values of an existing Nordic Semiconductor
nRF9160 LTE-M/NB-IoT modem [13] and thus they are
expected to reflect the values of an upcoming DECT-2020 NR
modem also.
TABLE III. ENERGY PARAMETERS
Parameter
Value
Battery capacity
18 kJ (5 Wh)
Voltage
3.7 V
Sleep state current
8𝜇A [15]
Rx/LBT state current
45 mA [15]
Tx state current
Linear Tx current = TxP (W) /
(3.7 V * 0.37) + 0.045 A [14]
The energy consumption of the nodes was collected only
after the initial network organization phase, so the increased
energy consumption of nodes due to additional beaconing and
RACH listening during the advertisement phase of the rotation
process was accounted for in the expected node lifetimes.
In the analysis, we assume minimum node battery life as the
metric to reflect the performance of the router rotation
protocol. Minimum node battery life is the shortest
experienced lifetime of a node in the network in the
simulations i.e. the maximum time all nodes are still in the
network, which is the desired situation. The longer the
minimum node battery life, the longer the network is intact
and the better is the performance of the rotation protocol.
V. SIMULATION RESULTS
Fig. 3 shows examples of simulated network topologies of
1000 nodes with some selected scenario radiuses. Different
network tiers are depicted with a different color. The figure
demonstrates how the network has only few tiers when nodes
are within a small radius, but as the nodes are dispersed over a
larger area, the network requires more tiers to cover the whole
scenario.
Fig. 3 Examples of simulated 1000 node network topologies with 800 m
(top left), 1200 m (top right), 2000 m (bottom left) and 5000 m (bottom right)
scenario radiuses.
Fig. 4 shows the minimum node battery life with different
user loads and scenario radiuses. With the given duty cycle
assumptions, without rotation, the first node dies after 0.1-1.5
years of network uptime, on average, depending on the load
and the mesh network radius.
With rotation, the average minimum lifetime is increased
threefold in smaller scenarios and low network load. The
rotation battery life gain increases as the load increases from
1000 to 3000 nodes but diminishes with a higher load. The
increase is likely caused by the increased number of candidate
routers: with a larger node pool to pick a new router from,
individual nodes can avoid the routing responsibility much
longer, prolonging their lifetimes.
Fig. 4 Minimum node battery life with different scenario radiuses and user
loads.
In absolute terms, router lifetimes decrease as more nodes
are introduced to the network due to increased traffic and
consequently increased activity in router nodes.
In larger scenarios the battery life gain from rotation is
smaller than with smaller scenarios. With larger radius the
network is spread thinner and there are less candidates near
router nodes to replace them. Thus the rotation responsibility
is rotated between fewer nodes and some nodes may even be
irreplaceable as routers, which decreases the overall
achievable gain in battery life.
Fig. 5 Percentage of nodes participating in traffic routing during simulation
with different scenario radiuses and user loads.
Fig. 5 shows the percentage of nodes who participate in
traffic routing at any given time during the simulations on
average. Without rotation, only a small percentage of nodes
act as routers in small scenarios. This is because in a small
geographical area, all the nodes can be covered easily with
only a few routers. In larger scenarios more nodes need to
participate in routing to achieve full coverage. Without
rotation, with the largest scenario radius, 20-40 % of network
nodes act as routers, which naturally limits the possibilities to
rotate their roles.
Fig. 6 shows the achieved end-to-end delays in different
cases. The delays are relatively long, measured in minutes in
all cases, but all being within accepted delay tolerance of the
use case. When comparing the results with and without
rotation, we see that rotation has a minimal effect on the
experienced delays. This is as expected since the rotation
interval is long in relation to the traffic generation interval.
Fig. 6 Box plots showing 5th, 25th, 50th, 75th and 95th percentile of
end-to-end delay with different scenario radiuses and user loads.
Also from Fig. 7, showing network service rate, we can see
that increasing user load significantly degrades service rate
due to data building up in the routers. Battery life is
consequently decreased due to increased activity from
retransmissions.
Rotation on the other hand degrades service rate so slightly
that if application QoS requirements demand it, the
degradation can most likely be counteracted with a minimal
RACH duration increase, which presumably would not
decrease the router lifetimes as much as the rotation would
increase it.
Fig. 7 Network service rate with different scenario radiuses and user loads.
VI. CONCLUSIONS
In this paper we showed that a periodic router rotation is
able to balance the energy consumption of DECT-2020 NR
mesh network nodes more evenly and thus significantly
increase the lifetime of a battery-operated IoT sensor network
without degrading the network performance, such as traffic
delay and network service rate. The gain in node router
lifetime is inversely proportional to mesh network size in
terms of both geographical area and number of nodes.
REFERENCES
[1] DECT-2020 New Radio (NR) Specifications, Release 1, 2021. [Online].
Available: https://www.etsi.org/committee/1394-dect.
[2] N. A. Pantazis, S. A. Nikolidakis and D. D. Vergados, "Energy-Efficient
Routing Protocols in Wireless Sensor Networks: A Survey," IEEE
Communications Surveys & Tutorials, vol. 15, no. 2, pp. 551-591, 2013,
doi: 10.1109/SURV.2012.062612.00084.
[3] S. Mamechaoui, F. Didi and G. Pujolle, “A survey on energy efficiency
for Wireless Mesh Network,” Intl. Journal of Computer Networks and
Communications, vol. 5, 2013, doi: 10.5121/ijcnc.2013.5209.
[4] R. Kovalchukov, D. Moltchanov, J. Pirskanen, J. Säe, J. Numminen, Y.
Koucheryavy and M. Valkama, “DECT-2020 New Radio: The next step
towards 5G massive machine-type communications,” arXiv:2101.07158,
2021.
[5] A. Anttonen, P. Karhula, M. Lasanen and M. Majanen, “Enabling
massive machine type communications with DECT-2020 standard: A
System-Level Performance Study,” VTT Technical Research Centre of
Finland Research Report No. VTT-R-00367-21, 2021.
[6] V. Dhanwani, N. Kumar, A. K. Bachkaniwala, D. Rawal and S. Kumar,
“Assessment of candidate technology ETSI: DECT-2020 New Radio,” in
Proc. 5GWF’20,2020, doi:10.1109/5GWF49715.2020.9221186.
[7] M. Penner, M. Nabeel and J. Peissig, "URLLC performance evaluation
of IMT-2020 candidate technology: DECT-2020 New Radio," in Proc.
VTC2021-Fall, pp. 1-7,
doi:10.1109/VTC2021-Fall52928.2021.9625340.
[8] T. Nihtilä and H. Berg, "Energy Consumption of DECT-2020 NR Mesh
Networks," in Proc. EuCNC/6G Summit, 2022, pp. 196-201, doi:
10.1109/EuCNC/6GSummit54941.2022.9815770.
[9] W. R. Heinzelman, A. Chandrakasan and H. Balakrishnan,
“Energy-efficient communication protocol for wireless microsensor
networks,” in Proc. HICSS’00, 2000, vol. 2, pp. 10,
doi:10.1109/HICSS.2000.926982.
[10] B. Chen, K. Jamieson, H. Balakrishnan and R. Morris, “Span: an
energy-efficient coordination algorithm for topology maintenance in ad
hoc wireless networks,” Wireless Networks, 2001, vol 8,
doi:10.1023/A:1016542229220.
[11] G. F. Riley and T. R. Henderson, “The ns-3 network simulator,” in
Modeling and Tools for Network Simulation, K. Wehrle, M. Günes and
J. Gross (ed.), pp. 15-34, Springer 2010, ISBN:978-3-642-12330-6.
[12] A. Pekar, J. Mocne, W. K. G. Seah and I. Zolotova, “Application
domain-based overview of IoT network traffic characteristics,” ACM
Computing Surveys, Vol. 53, Issue 4, pp. 1-33, July 2021,
doi:https://doi.org/10.1145/3399669.
[13] T. Zugno, M. Polese, N. Patriciello, B. Bojović, S. Lagen and M. Zorzi,
“Implementation of a spatial channel model for ns-3,” in Proc.
WNS3’20, 2020, doi:10.1145/3389400.3389401.
[14] F. I. Di Piazza, S. Mangione and I. Tinnirello, “On the effects of transmit
power control on the energy consumption of WiFi network cards,” in
Proc. QSHINE, 2009, doi:10.1007/978-3-642-10625-5_29.
[15] Nordic Semiconductor nRF9160 modem current consumption. [Online].
Available: https://infocenter.nordicsemi.com
