Implementation of a Low-Cost Vehicular VLC

System and CAN Bus Interface

Vicente Matus∗† , Cesar A. Azurdia-Meza† , Sandra Céspedes † , Pablo Ortega† ,
Samuel Montejo-Sánchez‡† , Javier Rojas† , Ismael Soto§
∗ IDeTIC-ULPGC, PCT Tafira, 35017, Las Palmas de Gran Canaria, Spain
† Departmentof Electrical Engineering, Universidad de Chile, 8370451, Santiago, Chile
‡ Department of Electricity, Universidad Tecnolgica Metropolitana, 8330378, Santiago, Chile
§ Department of Electrical Engineering, Universidad de Santiago de Chile, 9170124, Santiago, Chile

vmatus@idetic.eu, {cazurdia, scespedes, pabloortega}@ing.uchile.cl, smontejo@utem.cl,

jrojascatalan@ing.uchile.cl, ismael.soto@usach.cl

Abstract—Vehicular communication networks are one of

the essential technologies needed to implement intelligent
transportation systems in smart cities. In such networks,
available/licensed technologies still need to develop robust-
ness and resilience to support critical applications for
safety and efficiency. One novel technology, complementary
to radio-frequency, is visible light communications (VLC),
which has the potential of taking advantage of light emitting
diodes (LEDs) that are widely deployed in car lamps and
traffic lights. Control area network bus (CAN Bus), the
control network of all modern cars, can be read through
the on-board diagnostics (OBD) port available inside the
cabin. Data from sensors and actuators of the car can be
robustly acquired from the bus and can be shared to allow
the vehicular network cooperatively build knowledge of the
kinetic data of each car through beaconing. In this work, a Fig. 1. Basic architecture of vehicular networks, conformed by OBUs
low-cost VLC system based on white LED technology, which inside cars, and RSUs hosted by the vehicular infrastructure, that com-
is increasingly replacing halogen bulbs in cars headlamps, municate under V2V and V2I schemes.
is implemented and validated in laboratory conditions at
a distance up to 1.5 meters between the transmitter and
receiver. Further, the implementation of a low-cost CAN Bus The dedicated short-range communications (DSRC)
interface is shown. Its velocity data acquisition is validated standard and wireless access in vehicular environments
by testing the system against a global positioning system
(GPS) device. The proposed low-cost CAN Bus interface
(WAVE) family of standards for vehicular communications
achieved high reproducibility of the GPS estimations and are based on IEEE 802.11OCB, technologies based on
was validated with a 0.9979 Lin’s concordance correlation orthogonal frequency division multiplexing (OFDM) at the
coefficient. physical layer, in the 5.9 GHz band, and licensing up to
Keywords—Control area network bus (CAN Bus), software a bandwidth of 75 MHz depending on the country. Just
defined radios (SDRs), vehicular networks, visible light
as all other cases of wireless communications, vehicular
communications (VLCs).
networks face an important issue of spectral scarcity [5],
which has led to different approaches to solving the
I. I NTRODUCTION problem.
Vehicular communications networks are a particular The application of cognitive radio (CR) networking has
and challenging case of wireless communications, where been studied [6] to allow vehicular communications to
cars host mobile nodes called on-board units (OBUs) and take advantage of existing spectrum holes that can be used
streets host fixed nodes called road-side units (RSUs) [1]. opportunistically, without compromising the experience of
These devices communicate in inter-vehicular schemes the licensed users. Investigations by Eze, E. et al. in 2015
usually called vehicle-to-vehicle (V2V) communications, [7] and Pagadarai et al. in 2009 [8], characterized the
and in road-vehicle schemes, usually called vehicle-to- available spectrum in television broadcasting networks in
infrastructure (V2I) communications, as shown in Fig. 1. rural and urban areas, that opens an alternative for sparse
The nodes of a vehicular network share data; such as conditions. In [9], the authors proposed a framework to
car kinetic variables, control and automation instructions, temporarily add channels to improve a DSRC vehicular
and infotainment streams. The exchange of kinetic data is ad-hoc network (VANET) by the use of CR.
done by the transmission of beacons [2], [3], which allows Vehicular communications occur in a wide range of
OBUs and RSUs to cooperatively build the map of their proximity conditions, from sparse to super-dense (see Fig.
surroundings [4], forming the web needed to implement 1), and must work properly in all of them [5]. For the
intelligent transportation systems (ITS), that are a vast densest scenarios, bands of higher frequencies, such as
control network intended to improve road safety and traffic millimeter waves (mmW) [10] and visible light, which
efficiency [1]. have a higher directionality degree and a shorter range
than radio frequencies (RFs), can be considered. As shown Bus), which communicates to the computer of the car,
in [11], visible light communications (VLCs) work better then the information is transmitted with an LED to a
in heavy traffic conditions, where DSRC usually fails. A photo-diode (PD), and finally, the information is received.
robust and resilient technology for vehicular communi- The system then consists of two modules: (1) a VLC
cations networks should mix different technologies and transmitter (Tx) and receiver (Rx), based on phosphor-
use each in the best conditions. Table I summarizes the coated LED and based on p-i-n (p-type, intrinsic, n-type)
conditions and technologies discussed so far. PD (PIN PD) respectively, and (2), a vehicle CAN Bus
Modern vehicles include light emitting diode (LED) interface based on a universal protocol interpreter and
technology that could be used to deploy VLC in vehicular a micro-controller. To make the system as versatile as
communication. LEDs are usually built using Silicon, but possible in the prototyping stage, Tx and Rx are designed
as well as other elements like Gallium, and Nitrogen. based on software defined radios (SDRs). Note that SDRs
Chemical compounds formed by them emit light in very are only for prototyping purposes and a final product
narrow bands of a few tens of nanometers. For example, or implementation would be able to be built by using
an Aluminum-Indium-Gallium Phosphate (AlInGaP) LED simple and affordable communications modules instead
will emit red light approximately in the band of 620- of the SDRs. The final list of parts chosen to implement
650 nm [12]. To generate white color or some colors the modules of the system is shown in Table II. Next
that are not available as single compounds, coatings are subsections discus the implementation of each module.
used. Most modern cars use white phosphor-coated blue
LED technology in its headlamps, and red and yellow TABLE II
M AIN HARDWARE PIECES USED TO IMPLEMENT THE MODULES .
(non-coated) LEDs in tail-lamps. The coating has lower *: SIM800L INSTALLED FOR VEHICLE - TO - CLOUD (V2C)
switching rates than the LED itself. A single non-coated COMMUNICATIONS THROUGH GENERAL PACKET RADIO SERVICE
LED can be modulated at a 60 MHz bandwidth [13], while (GPRS)
a white phosphor-coated blue LED can be modulated by
Module of the System Hardware Used (cables excluded)
the order of 3 MHz without filtering, and up to 20 MHz 2 x National Instruments USRP 2922
if the blue component is optically isolated [14]. 2 x Ettus LFRX
2 x Ettus LFTX
VLC Module
1 x Thorlabs PDA36A
TABLE I 1 x Generic 10W LED Floodlight
C OMPLEMENTARY TECHNOLOGIES THAT CAN FORM WIRELESS LINKS 1 x Self-made Signal Combining Circuit
IN DIFFERENT TRAFFIC CONDITIONS
1 x ELM327 Bluetooth Dongle
1 x Arduino Nano
Traffic Conditions Proposed Technologies CAN Module
1 x HC-05 Bluetooth
Sparse RF (DSRC), CR 1 x SIM800L (unused*)
Dense RF (DSRC), CR, mmW 1 x Laptop (GNU Radio, Arduino IDE)
Super-dense mmW, VLCs Integration of modules
1 x Gb Ethernet Switch

The objective of this work is to integrate widely-

A. VLC Module
available white LED technology with communications
modules and vehicular kinetic data from obtained CAN The VLC module consists of a transmitter, connecting
Bus using low-cost hardware. Such development first a USRP NI2922 with a LED lamp, and a receiver, using
aims to implement a testbed that would ultimately allow a USRP NI2922 connected to a PD36A photo-detector.
developing specific hardware for VLC based vehicular The block diagram of the module is shown in Fig. 2 and
communications. This effort to join both VLCs and CAN its final implementation is shown in Fig. 3. The VLC
Bus technologies for vehicular communications is a new Tx and Rx subsystems are designed to work in base-
research approach that has the potential to release new band or low-frequency, as opposed to the default USRP
capabilities that OBU devices could include in the near scheme which is a pass-band scheme. In our previous
future. work [15], the modification of the USRP internal daughter-
The structure of the paper is organized as follows. boards was performed, and Ettus low-frequency cards
In Section II, the description and implementation of the LFTX and LFRX were installed. The range of operation
designed system blocks are shown. The experiments made of the oscillator within the Ettus low-frequency cards is
with the testbed are shown and discussed in Section III. from DC to 30Mhz.
Finally, conclusions are given in Section IV. Thorlabs PDA36A photo-detector, based on PIN PD,
works in the visible spectrum and ultraviolet (UV) and
II. P ROPOSED I NTEGRATED VLC S YSTEM infrared (IR) bands. Peak responsivity is reached at 960
In previous work [15], the modular design of the nm with a value of 6.5 A/W, whereas the best performance
system was presented. Whereas in this manuscript, the of the PD is in the near-IR spectrum.
main objective is to integrate widely-available white LED The LED’s original power source or driver was replaced
technology with communications hardware and vehicular by a self-made signal combining circuit (SCC). It sets a
control networks in order to implement a low-cost VLC fixed current to turn the LED on and a variable current
based vehicular communications system. The main idea to modulate the LED’s intensity in function of a signal
is that an on-board computer gathers kinetic information received from the communications blocks implemented in
of the vehicle using its control area network bus (CAN the SDR.
 Tx Rx
Signal Trans
PIN
Combining LED Impedance
Biased Light Beam PD Photo-
Circuit Amplifier
Comm. current
Signal
Amplified
Communication Photocurrent
Signal Data Data
Sent Received
Computer
Software Defined Radio Software Defined Radio
(Baseband or Low Freq.) (Baseband or Low Freq.)

Fig. 2. VLC module blocks for testing transmitter and receiver.

Fig. 4. Implemented Signal Combining Circuit (SCC).

(Tx) Modified GbEth. Switch (Rx) Modified
USRP USRP

The ELM327 is one of the lowest-cost and widely avail-

able circuits, whereas the Comma.ai Panda has the most
Tx-Rx distance (d) advanced hardware, designed to implement self-driving
cars.
Thorlabs SM1 Tube The CAN Module of the proposed system consists
SCC + 10W LED Thorlabs PDA36A of an ELM327 based OBD interface and an Arduino
Nano micro-controller. The systems are communicated via
Fig. 3. A laboratory setup of the VLC module. Bluetooth, as shown in Fig. 5. Figure 6 shows the final
construction of the CAN module.

The electric design of the SCC was based on the Vehicle CAN Module
CANHIGH
topology proposed by [16]. The final schematic is shown Vehicle’s
CAN ELM327
Sensors and CANLOW OBD Port Computer
Bus CAN Interpreter
in Fig. 4. The circuit has two blocks or sub-circuits: (1) Actuators Data

Vehice’s
a fixed-current source based on a MOSFET controlled by Bluetooth Arduino
Sensors and
Link (HC-05) Nano MCU
Actuators Data
an OPAMP, and (2) a variable-current source based on a
BJT whose current is set by the communication signal.
The DC block of the SCC can also act as a dimmer of the Fig. 5. CAN platform subsystem capable of reading the data of the CAN
Bus.
LED by changing the input of the OPAMP Vref , then the
fixed-current Ibias will be
Heatsink
Vref OBD
Ibias = . (1) 9VDC Regulators ELM327
Rbias Interface
DC 11-14[V] SIM800L
The variable-current Isignal changes in function of the from car * GSM *
communication signal Vin . Assuming the the transistor is
operating in the active region and its collector current is
approximately equal to its emitter current (i.e. the current Arduino
gain α = β/(β + 1) is unitary), Isignal is given by Nano
(USB)

Vsignal − 0.7[V ] Vin + VRDC R2

1 +R2
− 0.7[V ] HC05 Bluetooth *: Unused. For V2C communications.
Isignal ≈ = .
Rsignal Rsignal
(2) Fig. 6. CAN module implementation using Arduino Nano, HC05 Blue-
tooth circuit, SIM800L GSM circuit (for vehicle-to-cloud communication
purposes), and an OBD interface based on ELM327 circuit.
B. CAN Module
The CAN Bus is a bus-architecture (shared medium)
network existing inside modern vehicles where all the C. Analog VLC System
digital sensors and actuators (or control units) interact. The implementation of the software for the computer
It is a family of communication protocols released by of the system was done in a Debian operating system
the International Organization for Standardization (ISO) running GNU Radio Companion (GRC) with Ettus USRP
and the Society of Automotive Engineers (SAE). CAN Hardware Driver (UHD). A wide-band frequency modula-
reading allows a controller to gather kinetic data directly tion (WBFM) was implemented to transmit sound (wav
from the vehicle’s sensors, taking advantage of existing files or the computer’s microphone input) through the
infrastructure and sparing the installation of new sensors. VLC channel and listened at the receiver side using a
To read the CAN Bus, an interface for the on-board speaker, as illustrated in Fig. 7. This configuration works
diagnostics (OBD) port of the car is needed. The alterna- as a demonstration of the system capabilities to transfer
tives for such interface found so far in this work are the information via visible light since the performance can be
following: OpenXC Vehicle Interface, Comma.ai Panda, judged (qualitatively) by the ear. To quantitatively evaluate
ELM327 based dongles and circuits, STN1110 based don- the performance of the system, .wav input and output files
gles and circuits, and MCP2515-MCP2551 based circuits. can be saved, and later compared using digital techniques.
 Optics
WBFM SDR SDR WBFM De-
(Protecting
Data (D/AC, and Optics (2 MHz Data

+
Stream Modulator 2MHz
Glass,
(Dark Tube) Carrier, and
modulator Stream
(Audio) (Δf = 75kHz) Carrier)
Reflecting Noise (Δf = 75kHz) (Audio)
A/DC)
LED+SCC Cone) (Dark Room, PD+TIA
Multipath)

Fig. 7. Block diagram of the software developed in GNU Radio for the SDRs along with the VLC hardware implemented.

60
TABLE III
Measurements
PARAMETERS OF THE A NALOG C OMMUNICATION E XPERIMENT
Linear Regression

Velocity acquired by CAN Module [km/h]

Parameter Value Unit 50 Identity Line
Tx Gain 0 [dB]
Rx Gain 30* [dB]
FM Center Frequency 2M [Hz] 40
FM Frequency Max. Deviation 75k [Hz]
SDR 32bit-wide Sample Rate 1M [Samp./s]
Sound Card 16bit-wide Sample Rate 44.1k [Samp./s]
30
*: Rx Gain set using photodetector’s own amplifier.

20
Final codes of the software implementation are available
in the GitHub project’s repository1 .
10
III. R ESULTS AND D ISCUSSION
The VLC module was set to send a pure tone of 1 0
kHz at different distances between the transmitter and the 0 10 20 30 40 50 60
Velocity measured by LocoME OBU's GPS [km/h]
receiver. The parameters of this experiments are shown in
Table III. The demodulated signal was saved to observe the Fig. 9. Velocity data acquired with CAN module and estimated by the
distortion introduced by the distance. Normalized curves GPS of an Arada Systems LocoMate ME OBU.
are shown in Fig. 8. Note that the experiment was done
in total darkness and static conditions and the non-line-
of-sight (NLoS) signal was blocked. This is an idealized
scenario. The reflections from the street and other cars
surfaces, the presence of sunlight and other sources of measurements. The car was driven in an open area in
light, and the weather conditions have to be considered order to have the best GPS performance. The results of 191
for a realistic setup. measurements are shown in Fig. 9. The Lin’s concordance
correlation coefficient [17] between both measurements is
Sent (1kHz sine) 0.9979 with a 95% confidence interval between 0.9972 and
1
0.9984, denoting high reproducibility of the GPS velocity
0
estimation with the CAN module data acquisition.
1
Received at d = . 5m
Normalized Amplitude

1
The main differences between the OBU and the CAN
0
module measurements are that CAN data comes in integer
1
1 Received at d = 1m values, while the OBU has a resolution of one-hundredth
0
of km/h. Thus, the OBU can be more precise. On the other
hand, GPS devices usually fail when the satellite signal
1 Received at d = 1.5m
1 faces obstacles, like inside tunnels, whereas the CAN Bus
0 data is designed to always be available since the control
1 of the vehicle depends on it. Therefore, the use of kinetic
0 2 4 6 8 10
time [ms] data read from the CAN module and transmitted via VLC,
as proposed by the system described in this work, is a
Fig. 8. Comparison of 1kHz sine communicated through the VLC viable solution to guarantee the short-distance delivery of
module varying Tx-Rx distance d.
critical information in vehicular networks, especially when
In order to also validate the performance of the CAN the main communication technology (i.e., DSRC) fails.
module, it was mounted on a vehicle and operated si- Finally, as future work, the system will be evaluated in
multaneously with an OBU model Arada LocoME with a more realistic scenario, in a dynamic environment, with
integrated GPS. Both platforms measured the velocity of the presence of other sources of light and reflections such
the car in motion by their own means: the CAN module as asphalt, car and traffic lights, sunlight, and different
acquired data from the CAN Bus of the car, whereas weather conditions. The proposed system could also be
the OBU made an estimation based on GPS location directly integrated with the vehicle’s CAN Bus. Further,
1 Github Repository link: https://www.github.com/LaboratorioTICs- the implemented systems should be arranged in an outdoor
UChile/CSNDSP-paper codes static setting.
 IV. C ONCLUSION [3] S. Bolufé, C. Azurdia-Meza, S. Céspedes, S. Montejo Sánchez,
R. Demo Souza, E. Fernández, and C. Estevez, “Dynamic Beacon-
The implementation of a low-cost VLC transmitting ing using Probability Density Functions in Cooperative Vehicular
system was done using a standard LED lamp and a self- Networks,” in Special Session on Resilient Smart city Transporta-
made electronic circuit known as SCC that modulates the tion - RESIST 2018, Within the 4th International Conference on
Vehicle Technology and Intelligent Transport Systems - VEHITS,
LED’s luminous intensity. The carrier of the modulated 2018.
signal was set at 2 MHz, which is in the typical range of a [4] M. Boban and P. M. d’Orey, “Exploring the Practical Limits
phosphor-coated LED switching rate capacity. The system of Cooperative Awareness in Vehicular Communications,” IEEE
Transactions on Vehicular Technology, vol. 65, pp. 3904–3916, 6
was implemented and validated in laboratory conditions at 2016.
a distance of up to 1.5 meters between the transmitter and [5] H. Hartenstein and K. P. Laberteaux, “A tutorial survey on vehic-
receiver. It was found that vehicular networks need to be ular ad hoc networks,” IEEE Communications Magazine, vol. 46,
pp. 164–171, 6 2008.
robust in order to support critical applications such as ITS. [6] J. Eze, S. Zhang, E. Liu, and E. Eze, “Cognitive radio technology
VLC systems are not capable of implementing robust links assisted vehicular ad-hoc networks (VANETs): Current status, chal-
alone because they fail in long distances. VLC systems lenges, and research trends,” in 2017 23rd International Conference
on Automation and Computing (ICAC), IEEE, 9 2017.
are indicated for a communication range of a few meters, [7] J. Eze, S. Zhang, E. Liu, E. E. Chinedum, and H. Q. Yu, “Cognitive
where other technologies find a lot of congestion. Thus, Radio Aided Internet of Vehicles (IoVs) for Improved Spectrum
Resource Allocation,” in 2015 IEEE International Conference on
complementing VLC with CR networking based on RF Computer and Information Technology, Ubiquitous Computing and
and mmW, and the existing DSRC standard could support Communications Dependable, Autonomic and Secure Computing,
the widest variety of cases. Pervasive Intelligence and Computing, IEEE, 10 2015.
[8] S. Pagadarai, A. M. Wyglinski, and R. Vuyyuru, “Characterization
In parallel, a low-cost CAN Bus interface was imple- of vacant UHF TV channels for vehicular dynamic spectrum
mented in order to have a way to gather kinetic data access,” in 2009 IEEE Vehicular Networking Conference (VNC),
from the vehicle without the need of implementing sensors IEEE, 10 2009.
[9] N. J. Kirsch and B. M. O’Connor, “Improving the performance of
in the car. The complementary use of the CAN module vehicular networks in high traffic density conditions with cognitive
and a GPS device can provide precise and reliable car radios,” in 2011 IEEE Intelligent Vehicles Symposium (IV), pp. 552–
velocity information, that is needed for DSRC beaconing 556, 6 2011.
[10] J. Choi, V. Va, N. Gonzalez-Prelcic, R. Daniels, C. R. Bhat,
and further for traffic control algorithms. The proposed and R. W. Heath, “Millimeter-Wave Vehicular Communication
low-cost CAN Bus interface achieved high reproducibility to Support Massive Automotive Sensing,” IEEE Communications
of the GPS estimations and was validated with a 0.9979 Magazine, vol. 54, pp. 160–167, Dec. 2016.
[11] A. M. Cailean, B. Cagneau, L. Chassagne, V. Popa, and M. Dimian,
Lin’s concordance correlation coefficient. “A survey on the usage of DSRC and VLC in communication-based
vehicle safety applications,” in Proceedings of the 2014 IEEE 21st
ACKNOWLEDGMENT Symposium on Communications and Vehicular Technology in the
The authors acknowledge the partial financial sup- BeNeLux, IEEE SCVT 2014, pp. 69–74, IEEE, 11 2014.
[12] Avago, “HLMP Precision Optical Performance Red, Green and
port of the Project RETRACT ERANet-LAC No. Blue 5mm Standard Oval LEDs Datasheet.”
ELAC2015/T10-0761, Project FONDECYT Iniciación [13] D. Tsonev, H. Chun, S. Rajbhandari, J. J. D. McKendry, S. Videv,
No. 11160517, Project FONDECYT Iniciación No. E. Gu, M. Haji, S. Watson, A. E. Kelly, G. Faulkner, M. D. Dawson,
H. Haas, and D. O’Brien, “A 3-Gb/s Single-LED OFDM-Based
11140045, FONDECYT Postdoctoral Grant No. 3170021, Wireless VLC Link Using a Gallium Nitride micro-LED,” IEEE
and FONDEF Grant No. IT17M10012, and the members Photonics Technology Letters, vol. 26, pp. 637–640, 4 2014.
of the Telecommunications, Astrophotonics, and SPEL [14] J. Grubor, S. C. J. Lee, K. D. Langer, T. Koonen, and J. W.
Walewski, “Wireless High-Speed Data Transmission with Phos-
Laboratories at UChile, and the GETIC Laboratory at phorescent White-Light LEDs,” in 33rd European Conference
USACH for their technical support. and Exhibition of Optical Communication - Post-Deadline Papers
(published 2008), pp. 1–2, 9 2007.
R EFERENCES [15] V. Matus, N. Maturana, C. Azurdia, S. Montejo-Sánchez, and
J. Rojas, “Hardware Design of a Prototyping Platform for Vehicular
[1] A. Maimaris and G. Papageorgiou, “A review of Intelligent Trans- VLC Using SDR and Exploiting Vehicles CAN Bus,” in First
portation Systems from a communications technology perspective,” South American Colloquium on Visible Light Communications 2017
in 2016 IEEE 19th International Conference on Intelligent Trans- (SACVLC2017), (Santiago, Chile), 11 2017.
portation Systems (ITSC), pp. 54–59, 11 2016. [16] Z. Ghassemlooy, L. N. Alves, S. Zvanovec, and M.-A. Khalighi,
[2] P. Ortega, S. Céspedes, S. Bolufé, and C. Azurdia-Meza, “Exper- Visible Light Communications: Theory and Applications. CRC
imental Evaluation of Adaptive Beaconing for Vehicular Commu- Press, 2017.
nications,” in 6th International Workshop on ADVANCEs in ICT [17] I. Lawrence and K. Lin, “A concordance correlation coefficient to
Infrastructures and Services, 2018. evaluate reproducibility,” Biometrics, pp. 255–268, 1989.