This document describes a methodology for virtual drive testing (VDT) of LTE-A vehicular environments as an alternative to physical drive tests. 3D ray tracing is used to generate dynamic vehicle-to-infrastructure channels along a test route in Bristol. Measured antenna patterns and an RF channel emulator are used to simulate the channels. Results from the VDT system show good alignment with measurements from physical drive tests and provide reliable and repeatable performance analysis without the costs of real-world testing.

1. Communication Systems & Networks
© CSN Group 2017
LTE-A Virtual Drive Testing for Vehicular
Environments
VTC 2017 Spring
Michael Charitos, Di Kong, Jue Cao, Denys Berkovskyy, Angelos A. Goulianos,
Tom Mizutani, Fai Tila, Geoffrey Hilton, Angela Doufexi and Andrew Nix
Communications Systems & Networks
University of Bristol, UK

2. Communication Systems & Networks
© CSN Group 2017
Contents
• Introduction
• Methodology
• Channel modelling and antenna radiation pattern
• System setup
• Key results
• Conclusion

3. Communication Systems & Networks
© CSN Group 2017
Introduction
• A Virtual Drive Testing (VDT) methodology for a MIMO LTE Vehicle
to Infrastructure (V2I) urban scenario is proposed and compared
with actual road drive tests.
• This unique and generic radio performance analysis process is
based on:
• 3D ray traced channel models;
• Theoretic or measured antenna patterns;
• RF channel emulation and hardware-in-the-loop radio measurements (BS and
On Board Unit (OBU)).
• VDT is shown to provide a reliable, cost efficient and repeatable
alternative to physical drive tests.

4. Communication Systems & Networks
© CSN Group 2017
Methodology
• Virtual base stations and vehicles are connected
as they drive around a digital 3D map of Bristol.
• 3D city-scale ray-tracing channel models are
used to generate dynamic V2I channels.
• Theoretic and/or measured antenna patterns are
incorporated by spatial and polarimetric
convolution with the channel data.
• The resulting channel are streamed into Keysight
PropSim F8 channel emulator, which is
programmed to generate the RF channels
between an LTE-A dual-BS emulator (a Rohde &
Schwarz CMW500) and a Samsung S5 mobile
client (representing the vehicular On-Board Unit).
Route Selection
Channel Generation
Data Logging
HIL VDT

5. Communication Systems & Networks
© CSN Group 2017
Channel modelling: ray-tracing tool
• Databases include 3-D terrain, buildings
and foliage; all represented at a spatial
resolution of 10m.
• Ray information along the drive routes
include: amplitude, delay and AoD/AoA
in the azimuth and elevation planes for
each MPC.
• Vehicle motion is modelled by adding
the Doppler-shift corresponding to each
MPC.
• YouTube animation of the radio channel
and handover between two BS:
https://youtu.be/tD6uyAFLm9U

6. Communication Systems & Networks
© CSN Group 2017
Channel modelling: antenna patterns
BS Antennas:
• Measured data for an 800MHz panel sector antenna
with a gain of 16dBi.
• BS uses three antenna panels (120° rotated) with a
downtilt of 10°.
On Board Unit (OBU) Antennas:
• Two antennas are housed in a single
shark-fin pod located towards the
rear of the vehicular rooftop.
• The patterns were supplied by JLR.

7. Communication Systems & Networks
© CSN Group 2017
Channel modelling: convolution
• Point-source 3D ray tracing was performed from the BS to each of the
vehicular OBU locations along the test route.
• The double-directional time-variant channel impulse response h for a
link is given by the following equation:
ℎ 𝑡, 𝜏, 𝛺 𝐴𝑜𝐷, 𝛺 𝐴𝑜𝐴 =
𝑙=1
𝐿
ℎ𝑙 𝑡, 𝜏, 𝛺 𝐴𝑜𝐷, 𝛺 𝐴𝑜𝐴
=
𝑙=1
𝐿
El 𝑡 𝛿 𝜏 − 𝜏𝑙 𝛿 𝛺 𝐴𝑜𝐷 − 𝛺 𝐴𝑜𝐷,𝑙 𝛿 𝛺 𝐴𝑜𝐴 − 𝛺 𝐴𝑜𝐴,𝑙
where
𝐸𝑙 𝑡 =
𝐸 𝑇𝑥
𝑉
𝐸 𝑇𝑥
𝐻
𝑇
𝑎𝑙
𝑉𝑉
𝑒 𝑗𝜑 𝑙
𝑉𝑉
𝑎𝑙
𝑉𝐻
𝑒 𝑗𝜑 𝑙
𝑉𝐻
𝑎𝑙
𝐻𝑉
𝑒 𝑗𝜑 𝑙
𝐻𝑉
𝑎𝑙
𝐻𝐻
𝑒 𝑗𝜑 𝑙
𝐻𝐻
𝐸 𝑅𝑥
𝑉
𝐸 𝑅𝑥
𝐻 𝑒 𝑗2𝜋𝜐 𝑙 𝑡

8. Communication Systems & Networks
© CSN Group 2017
Emulation system
• Channel emulator: Keysight PropSim F8
• BS emulator: Rhode and Schwarz CMW500
• OBU: Samsung Galaxy S5
• 2x2 MIMO with both downlink and uplink

9. Communication Systems & Networks
© CSN Group 2017
Route selection
• A representative route connecting two major points of interest in the centre of Bristol.
• 1809 ray tracing location points were chosen along the route, with the exact co-ordinates
based on the vehicular GPS logs from the real-world drive tests.
• A handover scenario was selected as vehicle moves between two BSs. Handover also
considered between different sectors.
OBU
BS 1
Ray-tracing between BS and OBU
BS2
BS1
Direction of Movement

10. Communication Systems & Networks
© CSN Group 2017
Test scenario settings
Environment Type Urban (Bristol city-centre)
Vehicle Route Length 2.7 km
Number of ray-tracing points 1809
Frequency of Operation 806 MHz (LTE band 20)
Channel Bandwidth 10MHz
BS antenna type Panel Antenna (16 dBi gain)
Vehicle antennas MIMO1 (6 dBi)
MIMO2 (8 dBi)
OBU Receive Sensitivity -95 dBm
Transmit Power 34 dBm
• The assumed BS transmit power was
calibrated based on the ray tracing
results and the logged data from the
real-world vehicular measurement
reports.
• Channel dependent link quality
metrics, such as Reference Signal
Received Power (RSRP) and Physical
Downlink Shared Channel (PDSCH)
throughput were logged from VDT
system and physical drive tests.

11. Communication Systems & Networks
© CSN Group 2017
Key results (1)
• Good alignment between the prediction and measurement (left: RSRP1;
right: RSRP2).
• Better alignment for RSRP1 since Transmit power was calibrated for the
primary antenna at the OBU.
• Handover was triggered based on OBU measured RSRP values, indicating
that the handover occurs in the region around user id 1400.

12. Communication Systems & Networks
© CSN Group 2017
Key results (2)
• The variability associated with the VDT process is less than that observed in
the real world.
• The emulation results faithfully follow the measurement trends.
• Without prior knowledge, it is not possible to determine which results are real
and which are generated from VDT.
• Throughput depends strongly on channel conditions, resource allocation and
the link adaptation algorithm applied at the BS.
•

13. Communication Systems & Networks
© CSN Group 2017
Conclusions
• The laboratory based conductive test process developed and
reported in this paper was shown to be more reliable and repeatable
than real-world drive tests.
• Antenna optimisation (type and location on the car) can be
performed before physical prototypes of the antenna and/or vehicle
have been developed.
• VDT is able to provide automotive manufacturers with a powerful
and cost effective alternative to on-road testing.
• The use of city wide geographic databases allows a wide range of
operating environments to be considered, including rural as well as
urban routes.