Vous êtes sur la page 1sur 8

An in-vehicle wireless sensor network for heavy

vehicles
Dhasarathy Parthasarathy, Russ Whiton, Jonas Hagerskans, Tomas Gustafsson
Volvo Group Trucks Technology, Advanced Technology and Research
Gothenburg, Sweden
E-mail – dhasarathy.parthasarathy@volvo.com

Abstract—A practical method for the design of an in-vehicle and in some cases using apps on their phones to control simple
wireless sensor network operating in a truck is presented. The vehicle functions using Bluetooth. The introduction of Tire
network has been dimensioned using promising candidate sensor Pressure Monitoring Systems (TPMS) [1] is in many ways a
applications that were chosen based on a set of key criteria game changer in this area. Legislation was introduced in the
including safety, security, and timing-criticality and commercial US mandating the presence of a system that monitors the
viability. An extensive set of experiments were carried out to pressure in all the tires of a car, which helps prevent dangerous
determine characteristics, feasibility and ease of integration of tire blowouts and assists in reducing tire wear and improving
such a network into the vehicle electrical system. A network of 10 fuel efficiency. The system has typically been realized by
node positions and 3 possible gateway positions was used as a
embedding a pressure sensor, either on the rim of the wheel or
reference platform. Initial results indicate that for truck variants
on wall of the tire (the direct TPMS) which communicates with
with up to 5 axles, there is near universal coverage using 2.4 GHz
IEEE 802.15.4 radios with link budgets above 100 dB. Links an Electronic Control Unit (ECU) in the vehicle. The system
between sensors and all gateway positions have been found to gathers pressure data from all the wheels in the vehicle and
support above 90% packet reception rates at received signal displays the pressure values to the driver, along with a warning
strengths of about -70 dBm, with gateway positions in the central if boundary conditions are exceeded. The TPMS is a good
and rear chassis areas performing best. Current estimates are example of a feature whose implementation is made easier by
that such a network is robust enough to support sensors of low the usage of short range wireless communication. While this is
safety, security and timing criticality. yet to open the floodgates, it is easy to imagine the long reach
and flexibility of wireless technologies for in-vehicle use.
Keywords—In-vehicle wireless sensors; automotive wireless
sensor network

I. INTRODUCTION
The introduction of wireless sensors in vehicles is expected
to bring numerous benefits including simplification of wiring
harnesses, reduced maintenance costs, reduction of component
variants and related quality and cost improvements. The impact
of such a network of in-vehicle wireless sensors on the vehicle
electrical system needs to be carefully analyzed and accounted
for in order to ensure sustainable operation of vehicle wireless
sensors. This analysis requires two main tracks to be addressed
(i) the design of a network of wireless sensors operating in an
in-vehicle environment and (ii) integration of the network of
wireless sensors into the vehicle electrical system. Since the
use of wireless sensors in vehicles is in its infancy, new ground
is being broken in both aspects. This article presents the first
set of results from research conducted at Volvo GTT in the
area of vehicular wireless sensors. While the work being Figure 1 - Tire Pressure Monitoring System (TPMS) in a heavy duty truck
undertaken is quite broad in scope, this article concentrates
primarily upon aspects of dimensioning a truck-based wireless Remote controls were among the earliest applications
sensor network and the associated link level considerations. implementing short range wireless technology in vehicles.
Wireless key fobs are standard issue for most trucks on the
II. CURRENT STATE OF SHORT RANGE WIRELESS DEVICES IN market, while the Volvo FH also includes a wireless Work
TRUCKS remote (Figure 2) [2]. In addition, many new proposals are
made with regularity, like the use of a short range wireless link
The usage of short range wireless technologies remains for plug-in electric vehicles [3] or alarm systems to prevent
restricted to a relatively small subset of applications in trucks. theft [4], etc. but they are yet to be inducted into mainstream
Examples include drivers integrating the audio of their phones vehicle electronic systems.
978-1-5090-1314-2/16/$31.00 ©2016 IEEE.
were chosen by evaluating sensors based on the following key
criteria.

Key criteria Key concern

Sensors used for safety-critical functions require


highly dependable wireless links, but the maturity
Safety of dependable low power short range wireless
communication remains to be proven for in-vehicle
use cases.

Sensors used in security-critical functions require


Figure 2 - ECS4 Work remote for a Volvo FH
highly robust and secure wireless links, but the
Security maturity of secure low power short range wireless
III. USE CASES FOR WIRELESS SENSORS IN TRUCKS communication remains to be proven for in-vehicle
use cases.
As described above, short range wireless technology has a
relatively short history of usage in commercial vehicles and has
been restricted to mostly niche applications. This is Sensors usually need be able to fulfill timing
requirements both with regards to start-up time,
unsurprising considering the automotive industry’s relative Timing
transition time between power states and latency.
unfamiliarity with this technology, the tough operational reliability
Strict timing requirements will place high demands
environments that components in the vehicle need to operate on the low power wireless link.
under and tough communication requirements that they have to
fulfill. It is however evident that short range wireless devices, For energy autonomous operation, the electrical
Energy
particularly wireless sensors, offer many opportunities, autonomy
demands on the wireless sensor, particularly power
including: consumption, must be within reasonable limits.

1. Lower maintenance costs and increased uptime of the


The rate at which the sensor element is sampled and
vehicle, since most sensor-related faults can be traced Sampling and the rate at which sensed information is sent has
to faults in the wiring harnesses or the connectors, update rates direct consequences on power consumption and
which a wireless sensor eliminates. timing criticality.

2. Fewer sensor variants leading to reduced overhead,


since variations in sensor pigtails and connector RF circuitry is liable to be degraded at high
operating temperatures and an in-vehicle sensor,
encoding require additional administration and Operating
particularly in the chassis, usually needs to operate
inventory environment
in tough mechanical, micro-climatic, and chemical
conditions.
3. Easier installation of sensors on the assembly line due
to the elimination of harness routing – a predominantly
manual operation – because a wireless sensor offers a Commercial
The gains in removing harnesses, connectors,
overhead, etc. should trump new costs in radio HW,
single point of installation. feasibility
batteries and processes
4. New possibilities for sensing, since there are areas of
the vehicle where physical and logistical difficulties in Table 1 - Key criteria for choosing wireless sensor application candidates
deploying a wire harness would usually preclude the
installation of a wired sensor. A key fact to note is the huge diversity that in-vehicle
sensors exhibit when it comes to most criteria. An excellent
Thoughts and ideas on in-vehicle sensor networks example would be timing criticality, where sensors operate
sporadically pop up in literature, of which [5], [6] and [7] are under timing conditions ranging from very critical to
good examples. Research on wireless sensors for in-vehicle completely non-critical. Considering the huge range of choices
applications tends to focus mainly on link reliability and timing available, choosing sensor applications which place relatively
criticality resulting in extensive MAC layer work. There is also simple requirements on the low-power wireless link is a
a noticeable focus on security, for example [8] and [9],
pragmatic strategy.
primarily because of high-profile demonstrations of
vulnerabilities in wireless networks and how they can be With such a guiding principle, an extensive analysis of in-
exploited. vehicle sensor applications was conducted, resulting in four
Extensive work has been conducted at Volvo to identify candidates. For reasons of confidentiality, this article does not
suitable candidate sensor applications. A typical sensor in a disclose their exact nature but will provide information that is
truck needs to satisfy a host of requirements, covering major pertinent to the current discussion. It must be noted that the
areas of operational concern, such as functionality, interfaces, choice of the four listed candidates was also driven by initial
wiring, reliability, operating environment, packaging, estimates of commercial feasibility. A summary of key
maintenance and environmental compatibility. Being faced requirements for the candidates can be found in Table 2.
with a plethora of requirements, wireless sensor candidates
Candidate 1 Candidate 2 Candidate 3 Candidate 4
Criteria
(C1) (C2) (C3) (C4)
Safety
ASIL QM* ASIL QM ASIL QM ASIL B
criticality
Security
Non-critical Non-critical Non-critical Non-critical
criticality
Timing Semi-
Non-critical Non-critical Non-critical
criticality critical
Sampling/ Figure 3 - IETF Minimal 6TSCH stack
Update <= 1 Hz <= 1 Hz <= 1Hz <= 10 Hz
rates
V. A PLATFORM FOR LINK-LEVEL CHARACTERIZATION
Table 2 - Summary of key requirements for chosen candidates

*ASIL – Automotive Safety Integrity Level as defined by ISO26262

The determination of safety criticality of a sensor has been


is based on a combination of traditional FMEA and more
modern Functional Safety analysis techniques standardized by
ISO 26262. Most of the chosen sensors have been initially
determined to be non-safety critical (ASIL Quality
Management level) with one exception which is rated as ASIL
B, i.e. one that needs to have a fault rate of 10 -7 per hour.
When it comes to security criticality, an initial determination
has been made based on severity of tampering with the
information provided by that particular sensor. A more
rigorous determination would involve treating them to well
defined methods of threat analysis, which is not a primary
focus of this work. The determination of timing criticality is a
combination of freshness of data, tolerable losses and the
whether the sensor is used for status monitoring or if it is part
of a feedback loop in a control system. Among the chosen
candidates, one (C4) is a borderline critical case. The
remaining candidates rate much lower on the criticality scale.
IV. CHOICE OF A WIRELESS SENSOR NETWORK STACK
The chosen communication stack in this work is IETF
6TISCH [10]. This is based on the following:
1. The choice of an IEEE 802.15.4 link layer is feasible
since the theoretical maximum data rate (at 2.4 GHz
with O-QPSK and DSSS) of 250 kbps that it offers is
well-suited for the update rates considered in Table 2.
Link-level overhead in 802.15.4 has also been
designed with low power consumption in mind,
critical for extending lifetimes of battery-powered Figure 4 - Deployment of the reference network in a Volvo FH 6x2 tractor
sensors. The option of using the time-slotted channel
hopping MAC protocol defined by the 802.15.4e The exercise of determining whether an in-vehicle wireless
amendment allows the evaluation of semi timing- sensor network can be successfully deployed to meet the
critical applications. The relatively sound acceptance operational conditions is primarily one of link-level
of IEEE 802.15.4 in Internet of Things industry is also characterization. In its basic form, link characterization
strategically sound from the perspective of involves the generation of a predictable traffic pattern between
interoperability. nodes placed in representative locations, and monitoring
2. The choice of an IP-based stack eases interoperability performance under the influence of key link-level parameters
and the possibility to integrate into vehicle networks provides a good understanding of the link.
trending towards IP. With the IoT market trending A Volvo FH 6x2 tractor has been chosen as the reference
towards IP, this choice opens up the possibility of IoT vehicle and a network comprising of 10 sensor nodes ( (C1 – 4
vendors supplying sensors to the automotive industry. units, C2 – 1 unit, C3 – 1 unit and C4 – 4 units), an example of
which is shown in Figure 5, were deployed in locations shown
Parameter Values
in Figure 4. Deployment locations are primarily in the chassis,
spanning both the frame and the wheel hubs. A key question Vehicle Volvo FH 6x2 tractor
to be considered is the number of gateway ECUs needed to Gateway Central chassis (Gateway 1), Rear chassis
connect the network and their optimal locations. Four possible positions (Gateway 2) and Front chassis (Gateway 3)
gateway positions have been chosen for evaluation, with three
of them in the chassis and one in the cab. TX power 2.5 dBm (TX3), -12dBm (TX2) and -24dBm
(TX1)
Interfering Ambient interference (generally low levels)
sources only with no deliberate external sources
introduced
Vehicle mode 1 through 6 as defined in the Volvo FH manual
[14]
Payload size 5 bytes
Channel chosen IEEE 802.15.4 channel 26 (center frequency
2480 MHz)

Figure 5 - An ingress protected JN5168 USB dongle deployed in the reference Number of ~1000 at a best effort rate per test
truck packets per test

In our test platform, the sensor node HW is based on the Maximum 5


allowed retries
NXP JN5168 Wireless MCU [11] connected via USB to a
laptop for both power and logging. The open source Contiki Table 3 - Test parameters for Round 1 vehicle modes test
operating system [12] serves as the software platform, , which
has been extended by the partners of the EIT-ICT project
Reliable IP for Channel Hopping Networks (RICH) [13], to
provide an IETF 6TiSCH stack. The design of test cases, the
development of the link characterization platform and analysis
of experiments has been carried out in close partnership with
NXP Semiconductors, an approach that has brought about a
robust exchange of domain knowledge while at the same time
enabling independent review of results.
VI. LINK CHARACTERIZATION TESTS
A. Round 1 of experiments
The first round of experiments involved network tests to
obtain empirical link-level statistics of packet transmissions in
a Volvo FH 6x2 tractor. Characterization tests were centered
on three variables – TX power, gateway position and the
vehicle mode which decides the parts of the truck’s electrical
system that are operational. For each combination of these
factors, the link quality was measured using two main metrics
– Link Quality Indicator (LQI) as defined by the general PHY
requirements in IEEE 802.15.4 – which is based on received
signal strength and the other – Packet Reception Ratio (PRR)
– which is traffic based. Link characterization experiments
presented in this article were conducted under ambient
Figure 6 - Round 1: variation of LQI and PRR according to vehicle modes of
conditions with no deliberate introduction of external operation
interference.
A commercial truck can be operated in different modes – By exchanging traffic between all 10 nodes and 3 gateway
for example, parked, living, driving, etc. which activate positions and measuring the average LQI and PRR of each
different subsets of the on-board electrical system. To ensure link, it was found that the average value of the metrics as seen
the robustness of operation of wireless sensors it is necessary by the gateways (Figure 6) are not significantly affected by the
to establish the effect of mode of operation on the link quality. mode of operation of the vehicle. This lack of dependency of
The first set of tests in Round 1 – details of which can be the vehicle mode on link quality helps simplify the design of
found in Table 3 - concentrated on this question. experiments. The bulk of link analysis was then conducted
based on tests conducted on the road with the FH 6x2 tractor
being driven on a combination of highway and rural roads
around the city of Gothenburg.
Parameter Values
Vehicle Volvo FH 6x2 tractor
Gateway Rear chassis (Gateway 1), Central chassis
positions (Gateway 2) and Cab (Gateway 3)
TX power 2.5 dBm, -12dBm and -24dBm
Interfering Ambient interference (generally low levels) only
sources with no deliberate external sources introduced
Vehicle mode 6 as defined in the Volvo FH manual [14]
Payload size 8 bytes
Channel chosen IEEE 802.15.4 channel 13 (2415 MHz) during Figure 8 – Round 1: Node connectivity with Gateway 1
single channel tests and all 16 2.4 GHz channels
in channel hopping tests
Traffic pattern The coordinator broadcasts 100 packets to all
per test nodes while each node sends100 packets to the
coordinator. Packets are sent at a rate of 10 Hz
with 0 retries
Table 4 - Test parameters for Round 2 drive test

Key to note is the relationship between the two main


metrics, the observed link quality and the measured rate of
packet reception. An LQI value of 0-255 translated to
Received Signal Strength Indication (RSSI) in dBm is
reported by the radio chipset. A joint distribution between
RSSI and PRR (Figure 7) shows that for RSSI > -80dBm, a
PRR > 80% is virtually guaranteed. This observation reflects
the nodes’ RF- and link-level performance in this environment
and is quite fundamental to the design of an in-vehicle
wireless sensor network. It should be noted that PRR and
RSSI calculated once per test for each direction of a link, i.e. Figure 9 - Round 1: Node connectivity with Gateway 2
over 100 transmitted packets.

Figure 10 - Round 1: Node connectivity with Gateway 3


Figure 7 - Round 1: Joint distribution - RSSI, PRR
However, it is also important to understand the capabilities
of the network from the perspective of a more controllable
parameter – like TX power. From the spread of PRR between
all nodes and the rear chassis gateway, as a function of TX
power (Figure 8), it can be seen that at TX power levels -12
and 2.5 dBm, most nodes can achieve PRRs of more than
80%. A similar pattern can be seen in the connectivity of
nodes with Gateway 2 (Figure 9). Gateway 3 however
presents a stark contrast with very high spreads in observed
PRR in most cases – a marked characteristic of link instability.
To better quantify connectivity, a determination was made of
the lowest TX power level at which a particular node achieves
a PRR >= 90% for 90% of the observations (Figure 11).
Mapping this result to the physical location of the nodes
(Figure 12), it can be seen that only 4 positions are able to
achieve a link with PRR >= 90% with all 3 gateways.
Considering the fact that all node positions are chassis
based and, as shown in Figure 10, have unstable links with
Gateway 3, it is fair to isolate how nodes connect with only
the two chassis gateways. As apparent from Figure 11 itself,
and in Figure 13 where it is mapped into physical locations, all
nodes are able to connect quite comfortably with both the
chassis gateways. Node 219 in the rear chassis member has
strongest connectivity achieving PRR>=90% at -24 dBm with
both chassis gateways. All other nodes are able to connect at
either -12 or -24 dBm with each gateway, except 214 and 216
which need 2.5 dBm to have a similarly robust link with one
of the gateways. Figure 12 - Round 1: Node connectivity with all 3 gateways with PRR >=
90%

Figure 11 - Round 1: TX power necessary for a PRR of at least 90%

The levels of connectivity observed show that it is feasible


to operate chassis-based non-critical wireless sensors with a
chassis-based gateway. It does, however, remain questionable
whether the link can be used for a semi-timing/safety critical
sensor application.

Figure 13 - Round 1: Node connectivity the 2 chassis gateways with PRR >=
90%
B. Round 2 of experiments
To further qualify results of Round 1, a second set of
experiments was carried out, but this time with a longer truck.
A stationary 5-axle Volvo FMX 10x6 tipper truck (Figure 15)
was chosen as the test platform. Sensor node and gateway
node positions were chosen along lines similar to those of
Round 1.
Parameter Values
Vehicle Volvo FMX 10x6 tipper
Gateway Rear chassis (Gateway 1), Central chassis
positions (Gateway 2) and Cab (Gateway 3)
TX power 2.5 dBm (TX3), -12dBm (TX2) and -24dBm
(TX1)
Interfering Ambient interference (generally low levels) only
sources with no deliberate external sources introduced
Vehicle mode Parked
Payload size 2 and 8 bytes
Channel chosen IEEE 802.15.4 channel 13 (2415 MHz) during
single channel tests and all 16 2.4 GHz channels
in channel hopping tests
Traffic pattern The coordinator broadcasts 100 packets to all
per test nodes while each node sends100 packets to the
coordinator. Packets are sent at a rate of 10 Hz
with 0 retries
Figure 15 - Round 2: Node connectivity the 2 chassis gateways with PRR >=
Table 5 - Test parameters for Round 2
90%

General trends observed in Round 1 in the 6x2 are found to C. Major conclusions
recur in the 10x6, with Gateways 1 and 2 in the rear and
central chassis areas having good levels of connectivity with From the three rounds of experiments performed, the
the nodes. There is a marked similarity in levels of following conclusions can be drawn about the link:
connectivity observed, with only one exception of a node
position (213) in the fifth axle not having a 90% PRR link 1. For 802.15.4 chassis-situated wireless sensors, with
with Gateway 2. gateway ECUs in the central and rear chassis areas,
average packet reception rates above to 90% can be
consistently observed even at received signal strengths
close to -70 dBm, a condition achievable in all node
locations. These results indicate that gateway ECUs
are best placed in either or both of these areas.
2. Respectable levels of PRR at low TX power levels of -
12dBm or even -24 dBm show that lifetimes of most
nodes can be extended by reducing transmission
power levels without fatally compromising
connectivity.
3. The mode of operation of the vehicle (including
engine idling) seems to have little effect on packet
reception rates.
D. Ongoing work
While most of the conclusions established above are
expected to hold, the effects stemming from attached
Figure 14 - Round 2: TX power necessary for a PRR of at least 90%
accessories such as trailers and body builder units are
currently being evaluated. Link performance in the presence of
2.4 GHz wireless interferers is another major area of Carnegie Mellon University, 2010.
evaluation.
[6] R. Matischek, T. Herndl, C. Grimm and J. Haase, "Real-
VII. CONCLUSIONS time wireless communication in automotive
Following the introduction of short range wireless devices applications," in Design, Automation & Test in Europe
such as the TPMS in trucks, we now believe that the stage has Conference & Exhibition (DATE), 2011.
been set to evaluate if this technology can be expanded to
other sensors. Initial experiments from our research have [7] T. ElBatt, C. Saraydar, M. Ames and T. Talty,
indicated that IEEE 802.15.4 based wireless sensor networks "Potential for Intra-Vehicle Wireless Automotive
operating in the 2.4 GHz band in a truck are sufficiently robust Sensor Networks," in IEEE Sarnoff Symposium, 2006.
to host non-safety and non-security critical wireless sensors
with low-to-medium time criticality and update rates less than [8] H. Lee, H. Tsai and O. Tonguz, "On the Security of
10 Hz. Links from nodes to a central gateway located either on Intra-Car Wireless Sensor Networks," in IEEE 70th
the central or rear chassis areas have been determined to Vehicular Technology Conference Fall (VTC 2009-
achieve at least 90% packet reception rates at received signal Fall), 2009.
strengths of -70 dBm and above in the absence of significant
interference. Work is currently being conducted to [9] I. Rouf, R. Miller, H. Mustafa, T. Taylor, S. Oh, W. Xu,
characterize the network under additional scenarios including M. Gruteser, W. Trappe and I. Seskar, "Security and
driving with trailers and realistic levels of interference. privacy vulnerabilities of in-car wireless networks: a tire
pressure monitoring system case study," in USENIX
ACKNOWLEDGMENT Security'10 Proceedings of the 19th USENIX conference
on Security, 2010.
The research from DEWI project (www.dewi-project.eu)
leading to these results has received funding from the
ARTEMIS Joint Undertaking under grant agreement n° [10] IETF, "IPv6 over the TSCH mode of IEEE 802.15.4e
(6tisch)," [Online]. Available:
621353 and from VINNOVA, Sweden’s innovation agency.
https://datatracker.ietf.org/wg/6tisch/charter/. [Accessed
The authors would also like to acknowledge the extensive 19 Jun 2016].
collaboration between Volvo GTT and NXP Semiconductors
in developing and executing the experiments, and analyzing [11] NXP Semiconductors, "JN5168," [Online]. Available:
the results. http://www.nxp.com/products/microcontrollers/product
_series/jn516x/JN5168.html. [Accessed 13 08 2015].
REFERENCES
[12] Contiki OS, "Contiki: The Open Source OS for the
Internet of Things," [Online]. Available:
[1] U.S. National Highway Traffic Safety Administration, http://www.contiki-os.org/. [Accessed 05 September
"Federal Motor Vehicle Safety Standards, Tire Pressure 2015].
Monitoring Systems," [Online]. Available:
http://www.nhtsa.gov/cars/rules/rulings/tpmsfinalrule.6/
tpmsfinalrule.6.html. [Accessed 27 Nov 2015]. [13] EIT-ICT-RICH, "EIT-ICT-RICH," [Online]. Available:
https://github.com/EIT-ICT-RICH/contiki. [Accessed 3
September 2015].
[2] Volvo Trucks, "ECS4 with Work Remote," [Online].
Available: http://www.volvotrucks.com/trucks/uk-
market/en-gb/trucks/volvo-fh-series/key- [14] Volvo Trucks, "Manual for the new FH," [Online].
features/Pages/ecs4-with-work-remote.aspx. Available:
http://www.volvogroup.com/SiteCollectionDocuments/
VGHQ/Volvo%20Group/Logisticsolutions/handling_ins
[3] SAE International, "SAE and ZigBee team on vehicle- tructions/Volvo_Trucks/Manual%20new%20FH%20ful
to-grid standard," 2011. [Online]. Available: l%20version.pdf. [Accessed 13 Jan 2016].
http://articles.sae.org/9456/.

[4] Datachassi AB, "Datachassi leads the way for safe


transports," [Online]. Available:
http://datachassi.com/wp-content/uploads/2014/07/DC-
Broschyr-Eng_2014.pdf. [Accessed 10 April 2016].

[5] H.-M. Tsai, "Intra-car Wireless Sensor Networks,"

Vous aimerez peut-être aussi