Autonomous driving is expected to become part of our society from 2020 onwards. In Europe, the Horizon 2020 project considers the further reduction of traffic fatalities during the current decade as one of the main targets, as it suggests with the slogan “The Decade of Action for Road Safety.” Any autonomous car will have a sophisticated vehicle-to-whatever (V2X) communication infrastructure that will ensure the monitoring and even control of these vehicles at any time and under any conditions.
LTE technology is already playing a crucial role in connected vehicles, mainly for navigation, telematics and infotainment applications. By the end of 2015, connected vehicle growth is expected to reach 10 million and will be around 69 million by 2020.1
Next Generation Mobile Networks (NGMN), the forum that specified the requirements for 4G and selected LTE as the preferred standard, announced in its 5G whitepaper that civil aviation will implement commercial connectivity services starting in 2020, and passenger services offered will comprise similar applications to those available on the ground. This technology, referred to as direct-air-to-ground (DA2G) communication is relayed on cellular networks independent of the altitude or speed.
Ericsson2 launched some tests on LTE under extreme conditions, which consisted of testing the connectivity of LTE dongles fitted on an aircraft flying fast at low altitude. They demonstrated a maximum downlink speed of 19 Mbps internet connection while flying at 700 km/h against quite challenging Doppler scenarios. The tests also proved that a seamless handover from one radio base station to the next was possible while flying at a speed of 500 km/h.
Similar to DA2G communication, new flavors of V2X or vehicle-to-vehicle (V2V) based on cellular networks could be considered, offering an alternative to 802.11 systems. Connectivity on cars should be capable of handling all possible applications that could be ongoing inside a car simultaneously: emails, chat, web browsing, video streaming or infrastructure communications.
Under this scope, on-board-diagnostic (OBD) devices will play a key role in the autonomous car. Using the cellular network, they can provide real-time information of the status of the car at the precise location it is encountered. This information could be sent to any device within the surrounding infrastructure or to a cloud server for real-time monitoring, therefore highly appropriate for V2X or V2V systems.
This article proposes a laboratory test-bed solution for testing the robustness and performance of V2X and fleet tracking systems after analyzing the different sources of fading and network handovers as the main disturbers of communication while driving. The lab-based environment should play a big part in the whole testing process, as it provides several benefits over open road testing.
Sources of Fading
Cellular chipsets integrated into OBDs within a vehicle moving at speeds around 100 km/h are heavily impacted by multipath propagation problems. This phenomenon can give rise to interference that can reduce the signal-to-noise ratio (SNR) and degrade the bit error rate (BER) for digital signals.
In most wireless communication systems, the transmitter antenna does not have a fully directive radiation diagram, which can make the same signal arrive at the receiver through different paths at different times and with different amplitudes. As a result, the overall signal coming into the receiver will be the composition of all these components. This resulting signal can be an augmented or an attenuated copy of the original source depending on whether the multiple components are constructively or destructively combined.
The different amplitude and phase of the signals arriving at the receiver will determine this behavior. In most cases, the quality of the combined signal at the receiver deteriorates. As an example, the received signal power may vary by as much as three or four orders of magnitude (30 or 40 dB) when the receiver is moved by only a fraction of a wavelength.
Radio propagation models start initially looking at propagation path loss in free space (~ 1/r2). This model looks just at the signal strength when transmitter and receiver are in line-of-sight (LOS) condition. However, this model is not appropriate for describing the journey of a signal through different obstructions such as buildings, tunnels, hills or trees. The so called “shadowing effect” needs to be taken into account.
On one hand, the large scale propagation models or long-term fading models, which predict the mean signal strength for an arbitrary transmitter-receiver separation distance, improve the accuracy of the free space model by adding a lognormal distribution onto the strength of the received signals in decibels. Some examples are Okumura, Hata and Lee models.
On the other hand, propagation models that characterize the rapid fluctuations of the received signal strength over very short travel distances (~λ) or short time durations (~s) are called small scale or fading models. These models apply perfectly to automotive applications where transmitter and receiver positions are constantly changing. Small scale fading considers two main problems.
The first is multipath signals inducing inter-symbol interference (ISI) and delay spread. In these circumstances, linear filters with time varying impulse responses are used for modeling the mobile radio channels. Here, the time variation is due to receiver motion in space. In reality, the filtering nature of the channel comes from the summation of amplitudes and delays of the multiple arriving waves at any instant in time.
In this context, the term “flat fading” is used to describe the scenario in which the bandwidth of the signal is smaller than the bandwidth of the channel and/or the delay spread is smaller than the symbol period. Frequency selective fading is the term used for the opposite situation. Normally, frequency selective fading is modeled as the sum of several flat fading channels with different delays.
Rayleigh fading is usually considered as a flat fading channel model whose amplitude response follows the Rayleigh distribution, which can be seen as the envelope of the sum of two quadrature Gaussian noise signals. This model is commonly used to describe the statistical time-varying nature of the received envelope of a flat fading signal, or as the envelope of an individual multipath component.
In a situation where there is a dominant stationary (non-fading) signal component present, such as a line-of-sight propagation path, the small-scale fading envelope distribution is considered to be Rician. The channel, in this case, adopts the name of Rice channel. This model, very similar to Rayleigh, differentiates itself by way of a strong dominant component.
Having a transmitted continuous wave s(t),
s(t) = cos ωct
If the signal s(t) goes through a Rician multipath channel, the received signal r(t) can be expressed as:
r(t) = L cos ωct + ΣNn=1 αn cos (ωct + Φn)
L is the amplitude of the line-of-sight component.
αn is the amplitude of the nth reflected wave and will follow a Rayleigh distribution.
Φn is the phase of the nth reflected wave.
n = 1 .. N identifies the number of reflected, scattered waves.
Using this mathematical model, Rayleigh fading can be recovered for L = 0. Figure 1 represents both models.
The Rician K-factor is defined as the ratio between the power of the dominant component over the average scattered power at the receiver. Normally, the K-factor is expressed in dB. For indoor channels with an unobstructed line-of-sight between transmit and receive antenna the K-factor often settles between 6 and 12 dB. Rayleigh fading applies for K = 0 (- infinity dB).
A V2V radio link can be modeled statistically as a Rician fading channel with large K-factor. In these situations, the delay spread is expected to be relatively small because reflections will occur mainly in the immediate vicinity between the transmitter and receiver antennas. In general, the propagation channel is modeled with the following items3
- A dominant component consisting of a direct line-of-sight wave plus a ground reflected wave
- A group of early reflected waves
- ISI caused by the multiple delays between waves.
The second problem relates to a V2V system, where a moving receiver induces fading effects (Doppler shift) for each arriving wave component. When a pure sinusoidal tone of frequency fc is transmitted, the spectrum shape of the received signal, called the Doppler spectrum, will cover the range fc-fm to fc+fm, where fm is Doppler shift. The Doppler spread is caused by this Doppler Shift and can be expressed as a filter with the transfer function shown in Figure 2.
Where the Doppler shift corresponds to the widely known formula:
Having understood this, fast fading is referred to as a scenario with a high Doppler spread (large fm caused by high speed) and channel variations faster than baseband variations. Slow fading is just the opposite. These concepts are represented in Figure 3.4
Average fade duration is defined as the average period of time for which the received signal is below a specified level R. Average fade duration primarily depends upon the speed of the mobile (v), and decreases as the maximum Doppler frequency fm becomes large (at high speeds).
All these models and concepts were put in place for laboratory measurements via a fading simulator capable of simulating several fading types such as constant phase, pure Doppler, Rayleigh or Rice model, which were applied depending on different parameters like moving speed or angle of arrival. This fading simulator can be also used for conformance following the 3GPP TS36.521 where performance tests are specified against different multipath fading propagation conditions and delay profiles such as extended pedestrian A model (EPA), extended typical urban model (ETU) or extended vehicular A model (EVA).
These multipath delay profiles map the relative power with the delay of any tap relative to the first. In most cases a multipath fading propagation condition will be defined by a combination of a multipath delay profile and a maximum Doppler frequency which can be either 5, 70 or 300 Hz. The 3GPP TS36.521 specification also considers a high speed train condition with three scenarios: open space, tunnel with leaky cable and tunnel with multiple antennas. Only the second scenario is studied as a fading propagation channel, in which a Rician model is used for only one tap, with a Rician factor K = 10 dB, a Doppler frequency of 1500 Hz and a top speed of 300 km/h.
Since the deployment of the first 2G systems, mobility has been one of the key targets taken into account when designing the network. The second and third generations of mobile communications were developed targeting mobility support up to 100 km/h.
However, when 4G was deployed, even though its OFDM access scheme is robust against multipath interference, mobility scenarios switching the connectivity from 4G to 3G/2G were subject to studies. Furthermore, 3G and 2G still have a far better coverage than 4G (mainly present in big urban areas), and so many handovers can be expected when using cellular communications inside a car.
Under this scope5 it has been demonstrated how a change from 4G to 3G in mobile broadband affects the active connections. Some results show that transmission control protocol (TCP) has a slower throughput growth after a handover, when compared to user datagram protocol (UDP), and can suffer from extended periods of inactivity, also referred to as stalling. Since TCP is one of the core protocols of the Internet Protocol suite, used mainly by world wide web applications, it is essential to check that the vehicle applications transmit all their data to the infrastructure or cloud services at any time.
A key test that any OBD system should pass is when the car is going through tunnels. In tunnels, a modem might lose connection to 4G while still having decent 3G coverage. Other times, even the 3G connection is lost. All these imaginable scenarios can be tested in the laboratory by using the proper instrumentation.
The lab-based test system proposed in this article is essential for exposing the car’s V2X infrastructure and infotainment system against a fully controlled test environment which will represent real-life situations a user might experience while driving.
It is important to understand the key benefit of a lab-based testing approach, which is a controlled environment. Field testing can be very difficult to address and should be left for a second stage in order to confirm what has been predicted in the lab. Any solution aimed at testing on the open road brings several challenges such as finding the exact signal conditions to put the car under, logistic difficulties or even the safety of the environment, especially when sending control commands to the car. Lab-based solutions can be much more efficient. Figure 4 shows the proposed solution.
In order to explain how the lab-based approach is built, it is first necessary to understand the simulated operation for OBD applications. Initially, the infotainment system in the car will normally gather the OBD information via CAN buses. The GPS signal will be then collected from the integrated GPS receiver in the car’s communications module. Finally, a combination of both location data and OBD data will be taken by the infotainment application and uploaded to the V2X server hosted in the cloud.
The V2X server will be responsible for distributing the arriving data to the fleet tracking nodes settled along the whole infrastructure and will also communicate back to the vehicle(s) for reporting any information or, even, for control purposes. For example, the V2X server would be constantly analyzing the vehicles’ reports in order to share back information about the traffic. Also, if the V2X server detects speeding of a certain vehicle it could send a command to the OBD in order to decrease it automatically.
In the lab-based testing environment, the CAN port at the infotainment system will be connected to an OBD simulator in order to provide arbitrary OBD data. The network simulator will be used for testing the correct functionality and response of the system against stressed conditions, such as high AWGN level, low BTS power, cell interference or mobility scenarios. Thanks to the fading simulator, the RF signal being outputted from the network simulator will be disrupted depending on different fading conditions.
The adaptation of the signal for the different fading conditions is done digitally over the I/Q data of the transceivers. In this case, the fading simulator is connected to the network simulator by a LVDS cable for the baseband data and a BNC cable for synchronization. Once the BTS parameters have been configured on the network simulator, the fading profile is selected on the fading simulator. The fading simulator can be also adjusted for multi-antenna channel models. Figure 5 shows the differences in the cellular signal when not using a fading simulator and when using it.
The network simulator would be set up for simulating two cells of the same or different technologies in order to test the OBD system under more advanced scenarios, such as handovers or cell interference, plus adding complex fading channels on top.
The GNSS simulator will simulate the trajectory of a vehicle. Even though these simulators also support the addition of fading profiles, AWGN and antenna diagram imperfections, those details have not been included in this article.
Based on this idea, Anritsu and the University of Hertfordshire started a collaboration for implementing the proposed system. The University of Hertfordshire provided the OBD simulator and V2X server, and Anritsu provided the test instruments.
The V2X server can be further developed for analyzing the performance and functionality of the system against the different RF configurations. By saving the location along with the OBD data before it is transmitted from the infotainment system, the V2X server would be able to check that the arrival of the information has been successful and the information delivered is correct. Then the application on the V2X server could determine whether to send back a command to the vehicle or distribute the received information to real-time displays for fleet tracking purposes.
In Figure 6, a capture of the real-time V2X server developed by the University of Hertfordshire shows the trajectory being followed by the vehicle with a detailed telemetry report at every single captured point. This application can also detect some events as the one appearing at the picture and labeled as “over-revving,” which could eventually trigger a control message back to the vehicle in order to stop this behavior.
This article has shown a laboratory-based testing system for V2X and infotainment systems relying on cellular networks, which overcomes the challenges that testing on the open road brings. The most common sources of fading have been explained as well as how a fading simulator can be used within a test setup. Mobility problems can be also stimulated and addressed by using a network simulator capable of simulating two independent cells simultaneously. Finally, an example from the University of Hertfordshire’s real-time V2X server has been shown as a proof of concept of the proposed system.
- Mind Commerce, “Connected Vehicles: Market and Forecast for LTE and Telematics Applications 2015 2 2020,”June 2015.
- Ericsson News Center (2012) “Ericsson Tests LTE in Extreme Conditions.” Available at: www.ericsson.com/news/121101-ericsson-tests-lte-in-extreme-conditions_244159017_c.
- Jean Paul Linnartz, Wireless Communication Reference website (2010). Available at: www.wirelesscommunication.nl/reference/chaptr03/ivhs2.htm.
- ShareTechNote website, “LTE Quick Reference.” Available at: www.sharetechnote.com/html/Handbook_LTE_Fading.html.
- Patrick Skevik, “Study of how handovers in mobile broadband affects TCP,” Master’s Thesis Spring 2015, Department of Informatics, University of Oslo.