You are on page 1of 4

428 IEEE WIRELESS COMMUNICATIONS LETTERS, VOL. 7, NO.

3, JUNE 2018

Measurement and Configuration of DSRC Radios for


Vehicle-to-Train (V2T) Safety-Critical Communications
Junsung Choi , Vuk Marojevic, Aakanksha Sharma, Biniyam Zewede, Randall Nealy,
Christopher R. Anderson, Senior Member, IEEE, Jared Withers,
and Carl B. Dietrich, Senior Member, IEEE

Abstract—Despite the rapid development of wireless tech-


nology, there has been little application of the technology to
improve railroad crossing safety. We present vehicle-to-railroad
channel characterization and dedicated short range communica-
tions (DSRC) performance results at railroad crossings in rural
and suburban environments. Our results show that an omnidi-
rectional antenna provides slightly better performance in rural
conditions. However, a bi-directional antenna increased warning
range by more than 200 m in suburban conditions. A proper
configuration of the DSRC radio provides reliable warning of an
approaching train for cars near a railroad crossing.
Index Terms—DSRC, V2T, channel sounder, propagation char-
acteristics. Fig. 1. V2T communication scenario.

I. I NTRODUCTION Given the limited availability of empirical results in real


ESPITE recent efforts by projects such as Operation environments, we designed Revise to Vehicle-to-Train (V2T)
D Lifesaver to install safety features at crossings, up to
80% of U.S. railroad grade crossings are still classified
communications architecture and conducted measurements to
evaluate the suitability of DSRC for effective early warning at
as unprotected, and have no lights, warnings, or crossing grade crossings. This letter analyzes different antenna config-
gates [1]. Moreover, for the year 2016, train-to-vehicle col- urations and transmission parameter settings for the DSRC
lisions resulted in nearly 95% of all reported railroad accident transmitter on the train. The transmission parameters are
fatalities [1]. related to speed limits of the road and proper stopping
For Intelligent Transportation Systems (ITS), DSRC [2] is distance. Unlike typical V2V or V2I communications, the
the dominant protocol recommended for Vehicle-to-Vehicle designed system transmits warning messages of an approach-
(V2V) and Vehicle-to-Infrastructure (V2I) communications. ing train that is operating on a known track. We per-
Enabled by features of DSRC, ITS are able to project and formed measurements using omnidirectional and bi-directional
provide an early warning of a potential collision [3]. antenna configurations to explore the trade-off between trans-
Few studies address train collision avoidance warning sys- mission range and antenna beamwidth, extending our analysis
tems using wireless technologies. The only description of such in [5]. The measurements were done under a variety of oper-
a system that we are aware of is a train-to-vehicle early warn- ational conditions with the goal of finding the preferred
ing system designed and managed for railroad crossings in configurations. The wireless propagation environment was
Australia [4]. characterized to confirm classification of each experiment
site as a rural or suburban environment. The measurements
Manuscript received September 25, 2017; revised November 20, 2017; took place at grade crossings of the Shenandoah Valley
accepted December 5, 2017. Date of publication December 11, 2017; date Railroad (SVRR), a Class 3 short-line railroad in the Staunton,
of current version June 19, 2018. This work was supported by the Federal VA area. We describe our measurements, data analysis, results,
Railroad Administration under Grant DTFR53-13-C-00064. The associate edi-
tor coordinating the review of this paper and approving it for publication was and conclusions in the remainder of this letter.
Y. Gao. (Corresponding author: Junsung Choi.)
J. Choi, V. Marojevic, A. Sharma, B. Zewede, R. Nealy, and
C. B. Dietrich are with the Bradley Department of Electrical and II. M EASUREMENT T YPES
Computer Engineering, Virginia Tech, Blacksburg, VA 24061 USA
(e-mail: choijs89@vt.edu; maroje@vt.edu; sharma93@vt.edu; benyz@vt.edu; A. V2T Communication Scenario
rnealy@vt.edu; cdietric@vt.edu).
C. R. Anderson is with the Wireless Measurements Group Electrical and The V2T communication scenario considered here com-
Computer Engineering Department, United States Naval Academy, Annapolis, prises an approaching locomotive, vehicles approaching the
MD 21402 USA (e-mail: canderso@usna.edu). crossing, and optional infrastructure transceiver near the cross-
J. Withers is with U.S. Department of Transportation, Federal Railroad
Administration, Washington, DC 20590 USA (e-mail: jared.withers@dot.gov). ing as shown in Fig. 1. The scenarios comprise: the direct
Digital Object Identifier 10.1109/LWC.2017.2781734 warning case and the indirect warning case.
2162-2345 c 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
CHOI et al.: MEASUREMENT AND CONFIGURATION OF DSRC RADIOS FOR V2T SAFETY-CRITICAL COMMUNICATIONS 429

VA. For these measurements, a DSRC On-Board Unit (OBU)


was installed on a locomotive engine and a DSRC OBU was
installed near each crossing in a location that represents an
approaching vehicle.

C. RF Propagation Characterization
Propagation characterization measurements were performed
to measure channel parameters for the area near the railroad
tracks. Broadband propagation measurements at the crossings
were performed in the 5.86-5.91 GHz band, which overlaps
with the 5.875-5.885 GHz DSRC band. We used a Direct-
Sequence Spread Spectrum (DSSS) channel sounder [7] with
length 2047 m-sequence spreading code. The m-sequence was
clocked at a rate of 25 MHz (50 MHz RF bandwidth), and
GPS disciplined oscillators provided a stable frequency ref-
erence. Both transmitter and receiver were software-defined
Fig. 2. (a) Elevation pattern of omnidirectional antenna, (b) horizontal pattern
radio implementations using GNU Radio software and Ettus
of omnidirectional antenna, (c) elevation pattern and (d) horizontal pattern of USRP B210 hardware. A DSSS transmitter was installed about
a directional antenna used in bi-directional antenna configuration. 7-8 m away from the each crossing, and a DSSS receiver
was installed on the locomotive. Post processing measurement
data allowed us to calculate the RMS Delay Spread (RDS),
In the direct warning case, the vehicle driver receives the path loss exponents, and the Ricean K-factors in order to evalu-
warning message directly from the train. Direct warning is ate the channel conditions for the tracks. We then used channel
preferred when the speeds of the locomotive and the vehicles parameters to categorize each crossing as rural or suburban.
are slow, e.g., less than 35 mph (15.56 m/s). For this scenario,
the nominal stopping distance for a vehicle would be 84 m and III. M EASUREMENT PARAMETERS
the driver would have less than 5.4 s to stop [6]. In addition, A. Test Site Locations
the direct warning case is applicable where the propagation
Crossing #1 (Fig. 3a) was classified as being in a rural
between the locomotive and the vehicle has a strong Line-of-
area because of the relatively open space around the tracks
Sight (LOS) component.
and lack of significant infrastructure. Crossing #2 (Fig. 4a) was
In the indirect warning case, the locomotive broadcasts
classified as being in a suburban environment due to a greater
warning messages to an infrastructure transceiver near
number of buildings and structures near the grade crossing.
the crossing and the infrastructure retransmits the messages
to the vehicles. In this case, the infrastructure acts as a relay.
B. Radio Station Settings
The indirect warning case will be applicable for high loco-
motive and vehicle speeds, e.g., 65 mph (28.89 m/s), nominal DSRC measurements—We used Cohda MK5 DSRC
vehicle stopping distance and time of 210 m and 7.3 s [6]. radios [8]. One DSRC OBU was installed on top of the
The minimum notification time before the projected time of EMD GP9 locomotive’s long hood immediately behind the
a potential collision must include warning recognition time, Engineer’s cab and operated on channel 174 (centered at
driver reaction time, and stopping time. Warning messages 5.87 GHz); the antenna tip height from the ground was
must be sent before or at the minimum notification time. The approximately 5 m.
boundary to decide whether to use the direct warning or the The DSRC OBUs on the road vehicles were located at each
indirect warning case is therefore based on the DSRC radio crossing as marked in Fig. 3a and Fig. 4a. For each of the test
performance, where proper warning time/stopping distance is sites, the locomotive engine made a pass through the measure-
related to the speed of the vehicle and the locomotive. ment region. The locomotive engine started 300-400 m before
The experiments reported here address the direct warning the crossing, accelerated to the desired test speed of 10 mph
use case. (4.44 m/s) (the maximum speed permitted due to the track reg-
ulations), maintained this speed through past the crossing and
then decelerated to a stop after it passed 300-400 m beyond
B. DSRC Performance Measurement Method the crossing. Two types of antennas were used on the train:
The performance metric we chose to analyze the DSRC omnidirectional and bi-directional (two directional antennas;
performance was the distance over which packets were suc- one facing toward the front and one facing toward the back of
cessfully received with a Packet Error Rate (PER) less than the train).
0.9. The transmitter-receiver distance values are calculated The roadside OBU was installed on the roof of a pickup
between the GPS location of the crossing and the instantaneous truck at a height of about 2 m. The truck was positioned about
GPS location of the locomotive. All experiments were con- 60-70 m away from the tracks.
ducted with a stationary vehicle and moving locomotive. The Two omnidirectional antennas were mounted on the roof
measurements were performed at two crossings near Staunton, of the truck, as the roadside OBU’s utilize antenna diversity
430 IEEE WIRELESS COMMUNICATIONS LETTERS, VOL. 7, NO. 3, JUNE 2018

Fig. 3. (a) Crossing #1 receiver position and measurement track, (b) 3D PDP for omnidirectional, (c) PER with relative distance for omnidirectional, (d) 3D
PDP for bi-directional, (e) PER with relative distance for bi-directional.

Fig. 4. (a) Crossing #2 receiver position and measurement track, (b) 3D PDP for omnidirectional, (c) PER with relative distance for omnidirectional, (d) 3D
PDP for bi-directional, (e) PER with relative distance for bi-directional.

TABLE I
H ARDWARE C ONFIGURATION At these speeds, the typical vehicle stopping distance for a wet
surface road is 27 m and the typical stopping time is 2.43 s [6].
We consider the minimum notification time as 15 s; the loco-
motive moves 67 m in 15 s with speed of 10 mph (4.44 m/s).
With this information, if the DSRC performance is satisfac-
tory within 67 m from the crossing, we can assume that the
vehicle driver can receive the warning within the notification
to improve performance. The antenna patterns are shown in time of 15 s for the direct warning use case.
Fig. 2; other hardware configurations are shown in Table I.
Propagation measurements—Data was recorded to disk in A. DSRC Data Analysis
intervals of 0.5 s. The transmitter was placed at the roadside We used the Basic-Safety-Message (BSM) application with
and the receivers were placed on top of the locomotive. The a packet length of 99 bytes which was continuously transmit-
transmitter used an omnidirectional antenna and the receivers ted at one packet every 0.05 s. We evaluated PER every 20 m.
used the same bi-directional and omnidirectional antennas that The average number of transmitted packets during the time
were used for the DSRC measurements. window was 94. We set the threshold for packet reception
as receiving at least 10 packets corresponding out of 94; if
IV. DATA A NALYSES fewer than 10 packets were received at this distance, we con-
For Crossing #1 and Crossing #2, the speed limit on the cluded that reliable communication was not achieved. When
roads that intersect the railroad track is 25 mph (11.11 m/s). PER equals 0.9, at least 10 packets were captured, and we
CHOI et al.: MEASUREMENT AND CONFIGURATION OF DSRC RADIOS FOR V2T SAFETY-CRITICAL COMMUNICATIONS 431

TABLE II
P ROPAGATION C HANNEL PARAMETERS FOR using DSRC communication systems, we measured the propa-
C ROSSING #1 AND C ROSSING #2 gation characteristics and DSRC performance in terms of PER
and evaluated whether these systems could provide sufficient
warning time to allow a driver to come to a complete stop
before reaching the grade crossing.
For the performance evaluation for the given crossing and
speed regulation; we considered a range of 67 m with respect
to the crossing. Within this zone, all modulation types used
with high power and a bi-directional antenna met the require-
ment, for both rural and suburban environments. For low
power, an omnidirectional locomotive-mounted antenna would
concluded the performance as satisfactory, representing the be suitable in a rural environment, whereas a bi-directional
minimum requirement for safety. antenna would be better suited for a suburban crossing.
Our architecture was designed to demonstrate a proof-
B. Propagation Data Analysis of-concept early warning system with receive-only vehicles.
Power Delay Profiles (PDPs) were created by cross- Therefore, our results do not take into account co-channel
correlating the received channel sounder signal with a copy interference or greater network traffic that would appear in
of the m-sequence. Given the 25 MHz chip rate, our multi- a fully mature DSRC network. A greater number of full-
path time resolution was 65 ns. The Ricean K-factors is the duplex DSRC-equipped vehicles also means an increase in
ratio of LOS or dominant path signal power to multipath sig- the number of potential relay nodes, potentially providing
nal power in each PDP. The RDS and path loss exponent were a greater overall robustness of the delivery of the early warning
calculated from the PDPs using standard techniques published message.
in the literature. All calculations used relative position of the
locomotive and crossings. ACKNOWLEDGMENT
The authors would like to thank Benjamin True and the
V. R ESULTS Shenandoah Valley Railroad staff for their help planning and
The results are plotted as a function of distance from the conducting the measurements.
crossing, where negative distances indicate the locomotive is
approaching the crossing. R EFERENCES
The propagation channel parameters for Crossing #1 and [1] Federal Railroad Administration. (Sep. 2017).
#2 are given in Table II. Both the Ricean K-factor and path FRA Office of Safety Analysis. [Online]. Available:
loss exponents for omnidirectional antenna values are consis- http://safetydata.fra.dot.gov/OfficeofSafety/Default.aspx
[2] A. Rostami et al., “Stability challenges and enhancements for vehicu-
tent with values for rural environments reported in [11]. Since lar channel congestion control approaches,” IEEE Trans. Intell. Transp.
the position of the DSSS transmitter is near the crossing, the Syst., vol. 17, no. 10, pp. 2935–2948, Oct. 2016.
surrounding obstacles had little effect on the measurements [3] X. Xiang, W. Qin, and B. Xiang, “Research on a DSRC-based rear-
end collision warning model,” IEEE Trans. Intell. Transp. Syst., vol. 15,
and the measured values of channel parameters are consistent no. 3, pp. 1054–1065, Jun. 2014.
with those observed in rural environments. [4] J. Singh et al., “Cooperative intelligent transport systems to improve
Crossing #2, which we classified as suburban, experienced safety at level crossing,” in Proc. 19th ITS World Congr., Vienna,
Austria, 2012.
higher RDS, Ricean K-factors, and path loss exponents. [5] X. Ma et al., “Analysis of directional antenna for railroad crossing safety
As expected, the advantage of a bi-directional antenna is applications,” in Proc. Consum. Commun. Netw. Conf., Las Vegas, NV,
that it provides a longer range for Crossing #2. The dis- USA, 2017, pp. 1–6.
[6] Virginia Department of Motor Vehicles. (Jan. 2017).
tance for which satisfactory performance was achieved is Virginia Driver’s Manual. [Online]. Available:
200 m longer than for an omnidirectional antenna. Because the http://www.dmv.state.va.us/webdoc/pdf/dmv39.pdf
tracks are almost straight near the crossing; the bidirectional [7] C. R. Anderson, “Design and implementation of an ultrabroadband
milli meter-wavelength vector sliding correlator channel sounder and in-
antenna provides good coverage to the road near the crossing. building multipath measurements at 2.5 & 60 GHz,” M.S. thesis, Dept.
The disadvantage of a bi-directional antenna can be observed Elect. Eng., Virginia Tech, Blacksburg, VA, USA, 2002.
in the Crossing #1 data – it exhibits poor performance when [8] Cohda Wireless. (Jan. 2017). MK5 OBU. [Online]. Available:
http://www.cohdawireless.com/solutions/hardware/mk5-obu/
operating in a NLOS communication condition. [9] Mobile Mark. (Mar. 2014). Broadband, Omni-Directional DSRC
5.9 GHz. [Online]. Available: https://www.mobilemark.com/wp-
VI. C ONCLUSION content/uploads/2015/04/antenna-spec-144-eco-5900.pdf
[10] Ventev. (2017). 4.9-6.1 GHz, 20.5-23 dBi Dual Polarized
This letter discussed the performance of a V2T communi- Outdoor Panel Antenna With N Connectors. [Online]. Available:
cation system to prevent train-to-vehicle collisions by warning http://www.ventevinfra.com/
[11] R. He et al., “Measurements and analysis of propagation channels in
drivers near unprotected railroad crossings of an approaching high-speed railway viaducts,” IEEE Trans. Wireless Commun., vol. 12,
train. To evaluate the feasibility of providing warning signals no. 2, pp. 794–805, Feb. 2013.

You might also like