Conference PaperPDF Available

Open Source 5G-NSA Network for Industry 4.0 Applications



Content may be subject to copyright.
Open Source 5G-NSA Network for Industry 4.0
Elizabeth Palacios-Morocho, Pablo Picazo-Mart´
ınez, Sa´
ul Inca and Jose F. Monserrat,
iTEAM Research Institute, Universitat Polit`
ecnica de Val`
encia, Spain
{mapamo3, pabpima1, sauin, jomondel}
Abstract—Industry 4.0 seeks the digitization and
interconnection of all production processes. Due to the
huge number of sensors, robots, machines and other devices that
need to be interconnected, the Fifth Generation (5G) becomes
an enabling technology. In order to address this challenge,
this paper presents the implementation procedure of a 5th
Generation 5G Non-NSA private network based on Open Air
Interface (OAI) and presents the advantages and limitations
of OAI that will be useful for academia and the industrial
sector. A real use case has been evaluated to demonstrate the
capabilities of the network, where an Automated Guided Vehicle
(AGV) is remotely controlled and monitored using an open
source 5G-NSA private network. This paper presents the key
steps of the network implementation and analyzes the results
measurement campaign. Indeed, a network created with OAI
allows evaluating the benefits of many smart manufacturing
solutions but with certain limitations in terms of bit rate, latency
and number of active users in the private network.
Index Terms—Industry 4.0, Open Air Interface,
Non-StandAlone 5G, Latency, Automated Guided Vehicle,
Remote Driving.
The industry is constantly evolving and is now on its way
to a new stage known as the fourth industrial revolution or
Industry 4.0. This revolution is a paradigm shift that consists of
the digitization and interconnection of all production processes
with the aim of increasing market supply and increasing
operational efficiency [1].
Nowadays, the communications of industrial process
devices are based Ethernet or fiber. However, new technologies
are being introduced, such as Time Sensitive Network (TSN)
and Open Platform Communications Unified Architecture
(OPC UA), which enable important requirements for
automation systems to be met. [2] However, due to the
dynamism of Industry 4.0, a shift from wire-based solutions
to high-speed and reliable wireless solutions is required to
make the deployment of industry components more flexible
[3]. A strong candidate to provide the wireless solution for
Industry 4.0 is the 5G mobile. New Radio (NR) is the new
radio access interface that has been developed to support the
new requirements of 5G. Within those requirements are the
three types of communications defined by the International
Telecommunications Union (ITU) for 5G: Ultra-Reliable
and Low Latency Communications (URLLC), Enhanced
Mobile Broadband (eMBB) and massive Machine Type
Communications (mMTC) [4]. On the other hand, it is
important to note that two solutions have been proposed
as strategies for the deployment of 5G networks. The first
one, known as Non-Standalone (NSA), is a transition to 5G
where the radio interface is 5G but the control functions
relies entirely on the Evolved Packet Core (EPC) of Fourth
Generation (4G) Long Term Evolution (LTE) network. The
second solution, known as Stand-Alone (SA), is a complete
stand-alone 5G network that integrates both NR and a new
Fifth Generation Core (5GC).
An interesting 5G network deployment option for the
industrial and research sector is the use of private 5G networks
operating in unlicensed bands. To this end, one alternative is
to implement a 5G network using OAI [5]. OAI is an open
source initiative that provides a Third Generation Partnership
Project (3GPP)-compliant reference implementations of key
elements of 5G Radio Access Network (RAN) and 5GC that
run on general purpose computing platforms together with
Software-Defined Radio (SDR) cards for the radio functions.
[6] It allows users to set up a compliant 5G network and
inter-operate with commercial entities [7].
This paper presents how to deploy a OAI-based 5G network
in a factory environment and highlights its main advantages
and limitations. On the other hand, the performance of the
Fifth Generation Non-Standalone (5G-NSA) network has been
evaluated in a real scenario, where an AGV is remotely
monitored through the 5G-NSA network. The AGV remote
driving system is based on an operating system based on
Ubuntu LTS with open source licence called Robot Operating
System (ROS). The remote control of an AGV within
the industry allows optimizing the logistics in the plant,
performing tasks such as collection and distribution of all types
of materials. It also helps to reduce accidents in the workplace,
as it can be used to inspect high-risk areas for the operator.
Finally, for this testbed, an analysis of some KPI, such as
Signal to Interference plus Noise Ratio (SINR), latency and
throughput, is presented.
The remainder of the paper is as follows: in Section II
explains the 5G-NSA system setup using OAI. In Section
III the features of the AGV system setup base on ROS is
presented. In Section IV presents the discussion of the results,
in which an evaluation of coverage, throughput and latency is
made. The conclusions and future lines are found in section V.
This section presents the deployment of an open source
5G-NSA network based on OAI. The deployed configuration
is shown in Fig. 1, which consists of one server for the EPC
and two Universal Software Radio Peripheral (USRP) that are
configured as g-Node B (gNB) and Evolved Node B (eNB).
In this setup the EPC is used to launch 5G services and the
traffic is split for both the 4G and the 5G networks.
Fig. 1. Components used for the deployment of a OAI 5G-NSA network.
A. Technical Features
The EPC implementation is done by dockers on Ubuntu
18.04 installed on a server. The server contains the network
elements corresponding to the Mobility Management Entity
(MME), Home Subscriber Service (HSS), Serving Gateway
(S-GW) and Packet Data Network Gateway (P-GW). To
implement the HSS, Cassandra version 2.1 was used (available
in [8]). The same server is used for the implementation of
the RAN, for which the OAI repository has been cloned.
Additionally, two onmidirectional monopole antennas and two
USRPs have been used as radiating elements. It is important
to mention that the network operates in the NR 78 band, in
3.6 GHz. The technical features of the software and hardware
used are detailed in Table I.
B. USRP Setup
Before deploying the setup it is necessary to decide which
USRP model it is going to be used as radiance element. Ettus
Research provides a wide fan of possibilities, depending on the
use and performance desired. While serie “b” is cheaper and
simple to use, serie “N” provides better performance since
it can make bigger use of the specter. Serie “b” loads the
configuration in its Field Programmable Gate Array (FPGA)
each time an execution is made. On the contrary, serie “N”
includes a Linux based operating system that permanently
modify some of the features of the USRP in order to adapt it to
the desired deployment. An USRP makes use of his integrated
FPGA in order to perform all the radio functions via software.
Those functions were classically implemented using hardware,
like filters, transceivers or amplifiers. An FPGA is organized
in different sections each of which have a number of pins that
OAI Hardware
Component Features
Server Superserver 1029P-WTRT(Supermicro)
CPU Intel Xeon Silver 421632 cores @2.10 GHz
Hard Disk 960 GB SSD
Ethernet Ports 2x10 Gigabit Ethernet + 2x1Gigabit Ethernet
USB Ports USB 3.0
Radio device
Component Features
eNB USRP b210
Antenna Omnidirectional Monopole
OAI Software
Component Features
Operative System Ubuntu 18.04
Kernel Linux Low Latency
OAI Branch Develop
can be used to perform a particular function. In order to assign
a function to a FPGA pin, the USRP Hardware Driver Ultra
High Definition (UHD) is used. These drivers are adapted to
each USRP in order to simulate the radio features necessary
in order to allocate a 4G or 5G node.
During the deploy process of the 5G-NSA network within
an industry environment, two models of radiance elements
were used in order to make different tests. The first solution
includes 2USRP b210, a model that is designed for 4G
networks but can also work in 5G networks.
Despite the network was stable, its features were limited
by the hardware of the USRP, due to the low sampling rate
used, that only allowed bandwithds up to 40 MHz. In addition,
other features as using higher power or the possibility of
implementing Multiple-Input Multiple-Output (MIMO), are
not allowed using these devices.
Therefore, a second deployment was performed, in which an
additional bandwidth of 20 MHz was allocated to 5G, and the
gNB b210 was replaced by a higher-performance USRP, and
the latest USRP released by Ettus Reseach was used, the USRP
is N321. It can sample in a rate up to 200 MHz and perform
MIMO 2x2. Nevertheless, OAI solution was not compatible
with these USRP since the code worked with an UHD that
did not control USRP N321 FPGA properly. Consequently,
OAI code crashed, since the Linux machine was not able to
the control General Purpose Input/Output (GPIO) pins of the
USRPs motherboard. GPIO pins are used by OAI to exchange
information with the USRP, if those pins of the motherboard
are not operative, communication is not effective.
To make OAI code install this UHD, some changes in
the build were done. The script is called “build oai” and
is used by OAI to install all the dependencies necessary in
order to run their solution. In this script there is a function
that installs the the UHD is called “install usrp uhd driver”.
This function installs the UHD from the OAI repository and
they have the version 3.14 by default, which is not operative
with the USRP. To make the script install the 3.15 version
the UHD needs to be installed from source, which downloads
it from the provider Ettus Research. In order to make this
change, function “install usrp uhd driver from source” is
used. This function is located in the “build helper” script in
the path: “cmake targets/tools/build helper”. In the function,
“git checkout tags/v3.14” needs to be replaced with “git
checkout tags/v3.15.0.0”. Using these modifications, once
build is done with the command “build oai -I -w USRP −−
eNB −−gNB” it will install UHD 3.15.
Once UHD is installed properly in the Linux machine,
there are two ways of updating the UHD of the USRP. The
first one needs physical access to the device, and is done by
simply load by Secure Digital (SD) card the UHD version,
which was previously downloaded and moved to the card.
The second makes use of Mender in order to simplify the
process. The mender file contains all the processes needed
to update the UHD and needs to be sent using File Transfer
Protocol (FTP) to the USRP. Once this file is in the USRP,
Mender tool will remove previous versions and install the
desired one. If the version is not exactly the same as the
one installed in the Linux machine a similar error will pop:
“RuntimeError: FPGA component ‘noc shell’ is revision 5
and UHD supports revision w4. Please either upgrade UHD
(recommended) or downgrade the FPGA image”. This means
that some components of the UHD are not compatible within
Both the adaptations made to the OAI source code to make
it compatible with the USRP N321 and the eNB and gNB
configuration files that will be explained below are available
for be downloaded from GitHub repository at [9].
C. OAI Setup
The parameters set for the correct operation of the gNB and
eNB according to the EPC and USRP used in this deployment
are detailed in Table II. The same ones found inside the
configuration files for LTE soft modem and NR soft modem.
The Internet Protocol (IP) addresses assigned to the network
interfaces are configured in the range of our private network.
D. OAI Execution
The process to follow for the execution of eNB and gNB is
as follows:
1) Connect the first USRP to the server and run on the
$sudo./lte-softmodem- O<CONFIGURATION\
2) Connect the second USRP and run on the console:
$sudo./nr-softmodem- O<CONFIGURATION\
Centralized control of an AGV in an industrial environment
was developed with ROS Kinetic Kame, and interconnected
through a 5G-NSA network.
The interconnection of different entities is illustrated in the
box on the left-hand side of Fig. 2. The technical features of
Common parameters
Description Acronym Value
Mobile Country Code MCC 208
Mobile Network Code MNC 93
Tracking Area Code TAC 1
Number of antennas nb antennas tx 1
nb antennas rx 1
Resource blocks MN RB DL 100
Gain tx gain 90
rx gain 115
X2interface enable x2 yes
IP address of the MME mme ip address ”IP ADDRESS
X2interface enable x2 yes
LTE soft modem
Description Acronym Value
Band frequencies dowlink band 2680000000L
uplink frecuency offset 120000000
Networks interfaces ENB IPV4 “IP ADDRESS
NR soft modem
Description Acronym Value
Band frequencies absolute FrequencySSB 641272
dl frequencyBand 78
dl absolutyFrequency 640000
ul frequencyBand 78
X2interface target enb x2 ip address “IP ADDRESS
Networks interfaces GNB IPV4 ”GNB IP
X2C S1”
the software and hardware used for the deployment of an AGV
in an OAI network are detailed in Table III.
The architecture used to control the AGV through a
5G-NSA network is shown in the physical scenario depicted
in Fig. 2. The area of interest in the factory has 5rooms and a
corridor. The architecture follows a scheme client-server with
the client (Control Station) located in Room 1and the server
(Controlled Vehicle) located in Room 3. It is important to
mention that the tests was performed only in the rooms 13
(right-hand of Fig.2), which together have an area of 80 m2,
which corresponds to 10 m long and 8m wide. Room 1, with
an area of 6.25 m2, was assigned to the Control Station. The
other rooms represent the space in which the AGV will move
Fig. 2. Schematic of the AGV deployment and 5G-NSA network of the real scenario evaluated.
with a speed of 1.5m/s.
Room 1represents the control station where the factory
operator controls the movement of the AGV by means of
a steering wheel and visualises the environment through a
camera installed on the AGV.
The information about the current positioning, speed and
image is sent from the different sensors to the computer inside
the AGV. Then, via the router that has a Subscriber Identity
Module (SIM) configured with the parameters of our network,
the data is sent to the router in the control station. In order
for the data to pass through the network, the User Equipment
(UE) in this case the UE 1corresponding to the AGV must
connect to the gNB which in our case is the USRP N321 and
authenticate itself to the EPC. After successful authentication,
the data is routed to UE 2. In the opposite case, the positioning
and speed data are obtained from the sensor of the control
station (steering wheel) and sent from UE 2to the actuators
Component Features
PC Intel® Core™ i7-10750H CPUs @2.60 GHz (12
CPU Intel Xeon Silver 421632 cores 2.10 GHz
Camera USB 3.04 Intel RealSense Depth Camera D4351
AGV Summit robot
Router H112 370 “Huawei 5G”
Component Features
System Operative Ubuntu 16.04
Kernel Linux Low Latency
ROS Distribution Kinetic Kame
of UE 1.
The AGV is designed to work only indoors, can transport
small materials that together do not exceed 50 kg and has an
autonomy of 10 hour. In order to control it within an industrial
environment, it is necessary to transmit a good image quality
to the controller. In this scenario, a main stream resolution and
frame rate of 30 fps (1280 ×720) with a data transmission rate
of 7Mbps was set. It should be noted that this deployment
does not contemplate a solution based on planned routes in
which the AGV has autonomy, given that it is controlled by a
factory operator.
In order to analyze the coverage of the deployed network,
data were obtained at different points in the rooms 14. For
the room 5, an extrapolation was made with external points
as there was no access to it. On the other hand, the radio
equipment was located in the center of room 3, in order to
obtain the best possible coverage
A. Coverage evaluation
In order to evaluate the performance of the open source
network deployed in an industrial environment, an analysis
of the network coverage, Reference Signal Receive Power
(RSRP), Reference Signal Received Quality (RSRQ) and
SINR measurements has been performed. The SINR obtained
is shown in Fig. 3, where it can be seen that it reaches values
greater than 30 dB and minimum values close to 5dB. The
low SINR levels correspond to room 4 and are caused by
attenuations due to the walls and other obstacles. The first
conclusion is that the indoor propagation at 3.6 GHz is limited
and the range is not as good as it could be desirable.
Fig. 3. SINR Measurements.
In the coverage map represented in Fig. 4 the highest signal
power is found in the center of room 3, around the gNB and
has values near >70 dBm. The signal strength decreases
quickly while separating from the gNB and the coverage is
only good (meaning a value greater than 90 dBm only in
Room 3where the gNB was deployed.
Performing an analysis by rooms, Room 3has a SINR of
20 30 dB equivalent to excellent signal level. Rooms 12,
on the other hand, have a SINR of 10 20 dB which also
guarantees a good performance but the received signal power
presents an important degradation. Room 4experiences Signal
to Noise Ratio (SNR) levels less than 5dB. This room does
not represent an important factor for our final analysis since
the AGV will not be controlled in it. It is important to remark
that the SINR obtained is very high, due to the fact that the
network works in a free band, and it does not coexist with
other networks. This ismainly to the fact that the deployment
was carried out on the 1st floor of a 5-story industrial building
and the walls of the building almost totally diminish the power
Fig. 4. Coverage Map
Fig. 5. Bit rate Measurements
of the external signals. On the other hand, the deployment is
indoors, which helps to isolate it from possible interference
from outside, given the huge penetration losses of the walls at
the frequency of operation.
B. Throughput and latency evaluation
Although only a maximum bit rate of 228 Mbps in downlink
and 46 Mbps in uplink can be achieved as shown in Fig. 5, it
is sufficient for the control of the AGV, because between the
control data and the transmitted video a higher bit rate of 10
Mbps is required to achieve a correct operation.
Before discussing the results obtained on latency in the
OAI-based 5G-NSA network, it is necessary to mention
that the deployed solution theoretically allows to connect 16
users by default and up to 256 users if compiled using a
dedicated flag [10]. However, there were only 55G devices
available to perform the testbed. They managed to successfully
authenticate and connect to it, but only two of them can receive
resources at the same time. This is a important limitation of
the OAI-based deployment.
The latency experienced in the network in different
scenarios is shown in Fig. 6. In a first scenario where
only one user is connected, the minimum end-to-end latency
experienced is 19 ms. In other words, the time it takes to
transmit the data packet from the control station to the AGV
through the 5G-NSA network is 19 ms. As the user requests
more resources from the network, the latency increases to
values over 45 ms. In a second scenario where two users
(UE 1,UE 2) are requesting resources from the network at
the same time, the minimum latency raises a little but the
median increases from 26 to 28 ms. It is worth mentioning
that, in order to evaluate the difficulty of maneuverability,
of the AGV with different latency values, in addition to the
5G-NSA network described before, an additional OAI-based
LTE network was deployed. The deployment of the LTE
network is not discussed in this paper.
To evaluate the difficulty of maneuverability of the AGV
as a function of latency, a survey was performed in with 30
Fig. 6. Latency experience
people participated. They rated their experience on a Mean
Opinion Scale (MOS) of five degrees of difficulty. These were:
low, low-medium, medium, high and very high. The results
obtained are shown in the Table IV. It is observed that there
is a directly proportional relationship between the increase
in latency experienced in the network and the difficulty of
maneuverability of the AGV. Therefore, based on the results
of the tests carried out, it is recommended that the control of
the AGV be executed over a network whose latency does not
exceed 30 ms.
MOS Difficulty of
20 5 Low
25 5 Low
30 5 Low
40 4 Low-Medium
50 3 Medium
60 3 Medium
70 3 Medium
100 2 High
400 2 High
600 2 High
800 1 Very High
1000 1 Very High
This paper has shown an open source 5G-NSA network
deployment applied successfully to an Industry 4.0use case.
With values of SINR, in median, greater than 20 dB, good
signal quality and maneuverability was possible in the remote
driving of an AGV. Therefore, although the performance
of OAI is worse than that offered by commercial solutions
currently, it meets the coverage, stability and latency levels
required for the correct operation of the AGV communication
and control system in real time.
The stage of adapting USRP N321 to the OAI source code
has included bug fixing, register debugging and changes to
the OAI scripts, which are available in a GitHub repository
as indicated in the subsection II-C. In addition, the use of
N321 as a radiating element meant that higher bandwidths
were supported, which led to higher performance in the tests
carried out.
The 5G-NSA network based on OAI has enabled successful
and stable interaction between the AGV, the different
machines and operators. As a result, the time required to
perform tasks such as moving small tools from one place to
another has been reduced. Another advantage of using open
source networks is that it allows achieving a higher level of
debugging, since there is full access to all the source code.
Future research lines include the optimization of the
deployed 5G-NSA network, using the USRP N321 with
different MIMO configurations and a 5G Core Network. It
is also required to investigate other possible open source
solutions for 5G networks other than the one provided by
OAI, such as Open Ran (O-RAN) Alliance and to migrate the
centralized control of the AGV to a decentralized control, in
which navigation will be autonomous, using a real-time map
of the industrial environment.
Part of this work has been performed in the framework of
the H2020 project 5G-SMART co-funded by the EU. Saul
Inca and Jose F. Monserrat would like to acknowledge the
contributions of their colleagues from 5G-SMART although
the views expressed are those of the authors and do
not necessarily represent the views of the 5G-SMART
project. This was also supported by the Spanish Ministry of
Science, Innovation and University under the project RTI2018-
[1] J. Theunissen, H. Xu, R. Y. Zhong, and X. Xu, “Smart agv system
for manufacturing shopfloor in the context of industry 4.0,” in 2018
25th International Conference on Mechatronics and Machine Vision in
Practice (M2VIP). IEEE, 2018, pp. 1–6.
[2] M. Gundall, C. Huber, P. Rost, R. Halfmann, and H. D. Schotten,
“Integration of 5g with tsn as prerequisite for a highly flexible future
industrial automation: Time synchronization based on ieee 802.1as,
in IECON 2020 The 46th Annual Conference of the IEEE Industrial
Electronics Society, 2020, pp. 3823–3830.
[3] S. Tayade, P. Rost, A. Maeder, and H. D. Schotten, “Impact of short
blocklength coding on stability of an agv control system in industry 4.0,”
in 2020 IEEE International Conference on Communications Workshops
(ICC Workshops), 2020, pp. 1–6.
[4] Minimum requirements related to technical performance for IMT-2020
radio interface(s), Report ITU-R M.2410-0 Std., Dec,2017.
[5] H.-H. Cho, C.-F. Lai, T. K. Shih, and H.-C. Chao, “Integration of sdr
and sdn for 5g,” Ieee Access, vol. 2, pp. 1196–1204, 2014.
[6] F. Kaltenberger, A. P. Silva, A. Gosain, L. Wang, and T.-T. Nguyen,
“Openairinterface: Democratizing innovation in the 5g era,Computer
Networks, vol. 176, p. 107284, 2020.
[7] S. Martiradonna, A. Grassi, G. Piro, and G. Boggia, “5g-air-simulator:
An open-source tool modeling the 5g air interface,” Computer Networks,
vol. 173, p. 107151, 2020.
[8] Downloading cassandra. [Online]. Available: https://cassandra.apache.
[9] Downloading n321 code. [Online]. Available:
[10] number ue. [Online]. Available:
ResearchGate has not been able to resolve any citations for this publication.
OpenAirInterfaceTM (OAI) is an open-source project that implements the 3rd Generation Partnership Project (3GPP) technology on general purpose x86 computing hardware and Off-The-Shelf (COTS) Software Defined Radio (SDR) cards like the Universal Software Radio Peripheral (USRP). It makes it possible to deploy and operate a 4G Long-Term Evolution (LTE) network today and 5G New Radio (NR) networks in the future at a very low cost. Moreover, the open-source code can be adapted to different use cases and deployment and new functionality can be implemented, making it an ideal platform for both industrial and academic research. The OAI Software Alliance (OSA) is a non-profit consortium fostering a community of industrial as well as research contributors. It also developed the OAI public license which is an open source license that allows contributors to implement their own patented technology without having to relinquish their intellectual property rights. This new clause is in agreement with the Fair, Reasonable And Non-Discriminatory (FRAND) clause found in 3GPP. This paper describes the current OAI state-of-the-art of the development, the OAI community and development process, as well as the OAI public license and its usage by academia and industry.
5G-air-simulator is an open-source and event-driven tool modeling the key elements of the 5G air interface from a system-level perspective. It implements several network architectures with multiple cells and users, different mobility and application models, a calibrated link-to-system model for physical and data-link layers, and a wide range of features standardized for both control and user planes, as well as a set of technical components recently designed for the 5G air interface (such as massive Multiple Input Multiple Output, extended multicast and broadcast transmission schemes, predictor antennas, enhanced random access procedure, and NB-IoT). The tool has been already used in different research activities to design and evaluate the performance of reference 5G-enabled use cases. Moreover, it allows a flexible configuration, arrangement, and extension of its capabilities to model both new scenarios and new technical components, hence supporting advanced studies willing to address the research questions emerging from the deployment of current and upcoming mobile systems.
Wireless networks have evolved from 1G to 4G networks, allowing smart devices to become important tools in daily life. The 5G network is a revolutionary technology that can change consumers' Internet use habits, as it creates a truly wireless environment. It is faster, with better quality, and is more secure. Most importantly, users can truly use network services anytime, anywhere. With increasing demand, the use of bandwidth and frequency spectrum resources is beyond expectations. This paper found that the frequency spectrum and network information have considerable relevance; thus, spectrum utilization and channel flow interactions should be simultaneously considered. We considered that software defined radio (SDR) and software defined networks (SDNs) are the best solution. We propose a cross-layer architecture combining SDR and SDN characteristics. As the simulation evaluation results suggest, the proposed architecture can effectively use the frequency spectrum and considerably enhance network performance. Based on the results, suggestions are proposed for follow-up studies on the proposed architecture.