Professional Documents
Culture Documents
EXCLUSIVE:
Chris Borroni-Bird on AV Connectivity
mathworks.com/deeplearning
4 The Navigator
As the world turns to C-V2X, Europe picks WiFi
29 Hot New ‘Hands-On’ Direction in ADAS Safety Research New York, NY and additional offices) is published in March, May,
July, September, November by Tech Briefs Media Group, An SAE
International Company®, 261 Fifth Avenue, Suite 1901, New York,
JLR researches use of steering wheel heating and cooling to reduce NY 10016 and printed in Mechanicsburg, PA. Copyright © 2019
driver workload SAE International. Annual print subscription for SAE International
members: first subscription, $20 included in dues; additional
single copies, $30 each North America, $35 each overseas. Prices
30 5G Cellular-Enabled Test Infrastructure for AVs Launched for nonmember subscriptions are $100 North America, $150
overseas. POSTMASTER: Please send address changes to
at Millbrook Autonomous Vehicle Engineering, P. O. Box 3525, Northbrook, IL
60062. SAE International is not responsible for the accuracy of
Autonomous-vehicle 5G development support moves ahead at Millbrook information in the editorial, articles, and advertising sections of
this publication. Readers should independently evaluate the
Proving Ground in the U.K. accuracy of any statement in the editorial, articles, and advertising
sections of this publication that are important to him/her and rely
on his/her independent evaluation. For permission to reproduce
32 Delivering on Autonomous Transport or use content in other media, contact copyright@sae.org. To
purchase reprints, contact advertising@sae.org. Claims for
Einride ramps up industry partnerships and on-road demos of its all-electric missing issues of the magazine must be submitted within a
six-month time frame of the claimed issue’s publication date.
self-driving truck that’s missing one thing: a cab for the drivers. Address all communications for subscriptions or circulation to
Tech Briefs Media Group, Circulation Department, 261 Fifth
36 Upcoming, Ad Index
for a vaguely described commer- Argos of the world eventually bought by MARKETING Integrated Media
Joe Pramberger
cial-vehicle venture that will “integrate the auto industry’s established Tier 1s— Publisher Consultants
joe@techbriefs.com Angelo Danza
the Aurora Driver (hardware/software or acquired or otherwise absorbed by Debbie Rothwell +1.973.874.0271
adanza@techbriefs.com
platform) into FCA commercial vehicle the auto companies with whom they’re Marketing Director
drothwell@techbriefs.com Christian DeLalla
lines,” FCA’s brief statement said. What aligned. That scenario would leave Martha Tress +1.973.841.6035
Recruitment Sales Manager christiand@techbriefs.com
happened next? Former Aurora partner stack engineering as a de facto compe- +1.724.772.7155 Casey Hanson
Martha.Tress@sae.org
Volkswagen confirmed it was ending its tency of, well, the same companies that +1.973.841.6040
chanson@techbriefs.com
relationship with Aurora, immediately control most of the current automotive REGIONAL SALES Patrick Harvey
North America +1.973.409.4686
leading many to speculate VW will join supply chain. pharvey@techbriefs.com
New England/Eastern Canada:
Ford (with which it recently announced That may happen, but large sums ME, VT, NH, MA, RI, QC Todd Holtz
+1.973.545.2566
Ed Marecki
a commercial-vehicle partnership) in its will be involved. Is there a near-term +1.401.351.0274 tholtz@techbriefs.com
emarecki@techbriefs.com Rick Rosenberg
alliance with AV developer Argo AI. business case to support this calculated +1.973.545.2565
CT:
rrosenberg@techbriefs.com
The broad trend is evident in two and risky consolidation? Stan Greenfield
+1.203.938.2418 Scott Williams
respects: expect integrated AV devel- It’s a virtual certainty that today’s greenco@optonline.net +1.973.545.2464
Mid-Atlantic/Southeast/TX: swilliams@techbriefs.com
opers to increasingly align with, or 80-plus lidar developers will be whittled MD, DC, VA, WV, TN, NC, SC, GA,
acquire outright, key perception-tech to a handful. Will there be an indepen- FL, AL, MS, LA, AR, OK, TX SUBSCRIPTIONS
Ray Tompkins +1.866.354.1125
players. In turn, these developers are dent place for today’s AV tech integrators +1.281.313.1004
rayt@techbriefs.com
AVE@OMEDA.COM
firming their alliances with automakers. to co-exist with long-established Tier 1s NY, NJ, OH: REPRINTS
Ryan Beckman
Why? It’s becoming clear, even to the and the OEMs themselves? Billions of +1.973.409.4687 Jill Kaletha
rbeckman@techbriefs.com +1.574.347.4211
Waymos and Teslas of the world, that dollars rest on the outcome. I don’t have jkaletha@mossbergco.com
PA/DE:
(to highjack the AV industry’s well-trav- a clue how it will turn out; I only know the Desiree Stygar
+1.908.300.2539
eled phrase that originated with Tesla’s lineup card is getting smaller. dstygar@techbriefs.com
“ M ICH IGAN IS A
PLACE WHERE
PEOP LE G ET THING S
D ON E. ”
The state that revolutionized the way the world moves is doing it once
again. Across Michigan, talented engineers, technicians and scientists
are creating new mobility solutions, from autonomous vehicles to
connected infrastructure to cybersecurity. Mobility in Michigan is
moving fast, so get here or get left behind.
Visit planetm.com
AUG 21-23
Copyright © 2019 Dell Inc. or its subsidiaries. All Rights Reserved. Dell Technologies, Dell, EMC, Dell EMC and other trademarks are trademarks of Dell Inc.
or its subsidiaries. Other trademarks may be trademarks of their respective owners.
Free Info at http://info.hotims.com/73014-704
Software/Simulation
Complexity of
Autonomous-Systems
Simulation, Validation
Soars to the Clouds
by Terry Costlow
As the auto industry strives to improve safety that run operations in parallel to shorten analysis times.
and edge towards high-level automated driving, the Automakers and Tier 1s have acknowledged the
complexity of proving that electronic vehicle controls will value of simulation as an integral tool for getting the
perform safely is skyrocketing. Simulation’s expanding myriad software and hardware elements of automated
role in systems validation is prompting many tool driving to work in collaboration. High-level automated
providers to move to scalable, cloud-based architectures driving requires vehicle controls to “understand” many
complex external elements, which have far more
variables than conventional on-vehicle systems such
as engine controls. Cameras, radar and lidar must
monitor pedestrians, vehicles and highway markings.
Metamoto
tion that’s important for decision making. The lack of
information about outputs makes it difficult for simu-
Validating the many elements of urban traffic requires many thousands
of simulations.
lation suppliers to know whether they’re accurately
representing sensor outputs. Most sensor suppliers
have been hesitant about divulging the code they use—
the chips used in design, taking those chips and building but that stance may be changing.
systems and going to the city level,” Worm said. “At the “We’re interested in taking a given number of
city level, we’re a big supplier of infrastructure equipment.” scenes and seeing if our model gives us the same results
on the output side,” Blackburn said. “We’ve got formal
Test commonality is crucial and informal relationships with radar, camera and lidar
The complexity of testing many vehicle types in providers; we asked them to provide a scenario, then
multiple driving scenarios poses a major challenge we shipped them our raw data from that scenario and
for those tasked with ensuring AVs are safe. Many of asked them to compare it to their actual data. It’s an
those who are involved with development and valida- iterative process.”
tion believe that some common tests will be needed As with most developmental aspects of autono-
as a foundation to determine whether vehicle systems mous technologies, there are many subtleties that can
perform safely. alter both virtual and physical sensor outputs. Weather
“It’s important to do repeatable tests,” Worm conditions are a factor for cameras, for example.
said. “Everyone needs to do the same tests with Lidar’s addition of light bouncing off the target brings
different vehicles. a number of extra parameters.
The need for common tests extends to the “With lidar, something that’s sometimes missed is
critical regulatory aspect of vehicle certification. The that you need to know whether the beam is hitting a
National Highway Traffic Safety Admin. (NHTSA) wet or dry surface, whether it’s a pedestrian or metal,”
and industry are expected to work together to ensure Blackburn said. “The software needs to account for
that all vehicles are tested using the same criterion. material properties.”
Given the complexity of ensuring that vehicles respond Artificial intelligence will be an important
well in myriad situations, no set of tests can guarantee element, helping tool providers develop tests and
complete safety—but few standard tests may have to aiding system developers to improve their data-gath-
be used to demonstrate a base level of competence. ering and analysis techniques. AI can assist compa-
“If simulations are going to be used as part of the nies building scenarios for testing, helping them vary
vehicle validation process, I feel you have to have an conditions and create tests that might be difficult to
agreed-upon library of scenarios,” Blackburn said. “I run in the real world.
don’t think NHTSA can develop a testing technique When it’s used in vehicle systems, AI will pose
for autonomous driving—and it’s not like NHTSA will many challenges for those tasked with validating safety,
let automotive companies self-verify their vehicles.” since its response to stimuli can change depending on
conditions—meaning more potential variables. Once
Not all sensor outputs are equal more, it’s likely that companies will come together to
While regulators grapple with their aspects of vehicle create some standard scenarios that will test the tech-
safety, engineering teams are wrestling with the nology’s output to determine that it’s safe. ■
Connectivity
Solutions for AVs
by Chris Borroni-Bird
and has to be restarted from scratch. The update may functionality easier). OTA solutions are necessary for
even require that the EV is not charging at the same time securing the vehicle throughout its lifecycle because
and delaying charging may mean reduced vehicle range even secure embedded vehicle electrical and electronic
and the associated inconvenience. systems can become vulnerable over time to increas-
These issues, though troubling, are less worrying ingly sophisticated cyberattacks if there is no updating
than the second type of challenge: ensuring the of the security protections.
security of the vehicle and its associated cloud system
service while the update occurs. Cybersecurity risks Robotaxi deployment hurdles
can harm the vehicle systems and render vehicle Several companies are developing solutions that will
operation unsafe with loss of control and vulnera- enable robotaxi operation in an automated MaaS
bility to theft. Although non-connected vehicles are model. This is SAE Level 4 where the AV may have
susceptible to hacking, the probability and severity of no ability to be manually driven but can operate in a
hacking will be significantly increased with connected geo-fenced zone, such as part of a metropolitan area,
vehicles in the future, especially connected automated that has favorable regulatory and climate conditions
vehicles which will have many more attack surfaces. for AV deployment.
Fortunately, mobility-tech companies are developing The currently favored approach to developing
holistic approaches to end-to-end security between a robotaxi service is to do testing in the geo-fenced
vehicle and cloud. location environment with a safety driver who takes
OTA software updates for critical vehicle function- over control of the vehicle when the AV system is chal-
ality may need to be delivered from an OEM-controlled lenged to make the correct decision. This testing is
gateway to a dedicated modem on the vehicle, separate intended to allow the AV software to be improved by
from the customer-facing one needed for delivering learning how it fails in the real-world without jeopar-
Internet content to the vehicle’s infotainment system dizing safety for other road users. The work is comple-
(this need for cellular modems makes adding C-V2X mented with extensive simulation of conditions that
Continental Automotive
network-based communications between vehicles, to the infrastructure
and to vulnerable road users tested under varying conditions. The companies
observed a mean latency of 20 ms for direct communication and a nearly
error-free communication even at longer distances such as 1.2 km with
unobstructed line-of-sight condition as measured by the C-V2X test system.
may occur very rarely during the miles accumulated have been limited mainly to pilot programs and don’t
in physical testing. have the same scale as for equipped passenger vehicles.
What makes the development of an AV solution Since some cities are keen to promote the deploy-
even more challenging is that the vehicle not only ment of AVs because they promise to improve mobility
needs to drive safely without causing any collisions, but and safety, it is reasonable to imagine that funding
it also needs to behave in a manner similar to that of for I2V installations (e.g. vision-based sensors plus
human-driven vehicles (i.e. roadworthiness). It needs wireless communications) at those locations which are
to be able to infer the expected trajectories of other challenging for AVs could help to accelerate safe and
road users and to be able to proceed in ambiguous viable AV deployment in perhaps a more cost-effec-
situations and make necessary adjustments based on tive and timely manner than without such support. A
the actions of other road users. simple example might be that of an AV having trouble
It may be possible to achieve a viable performance quickly negotiating an intersection in a safe manner
using the AV’s embedded sensors and software, but as it climbs a hill. With the addition of vision-based
this has not been proven in practice so far. It will be sensors, such as camera, lidar or infrared camera,
argued here that connectivity with roadside infrastruc- information about presence of absence of cross-traffic
ture and with a Tele-Operator may allow a viable AV to and of potential violations can be wirelessly commu-
be developed sooner without jeopardizing road safety nicated to the AV and allow it to make a more robust
and traffic flow. and faster decision on how best to proceed.
Various reasons have been given for why V2V has The additional cost to the infrastructure and
not been considered integral to AV deployment. There vehicles in the fleet needs to be weighed against the
has been a lack of global (and even regional) agreement extra time and effort required to improve the software
on standards for DSRC and the emerging challenge it to a point that the same level of safety and traffic flow
faces from C-V2X has exacerbated this. The benefits is achieved. In this scenario, the AV fleet does not need
of V2V require large-scale deployment which will to communicate with other vehicles and the number
take several years to achieve. Meantime, there is no of roadside installations is finite and relatively small.
perceived business model incentivizing automakers If a city funds the roadside installation it is possible
to introduce it on their vehicles. Infrastructure-based that this could be used for the benefit of all road users
solutions for improving commercial vehicle operation and not just the AV fleet developer, whereas if the AV
“
Connectivity with roadside infrastructure and with a
Tele-Operator may allow a viable AV to be developed
”
sooner without jeopardizing road safety and traffic flow.
—Dr. Chris Borroni-Bird
fleet developer were to fund the installation this may it harder to see with monochrome images). Virtual
or may not be the case. As the number of vehicles in Reality Head-Mounted Displays (HMDs) tend to
an AV fleet increases there could also be value in each provide a more compelling immersion experience
AV being able to sense their surroundings and share for the Tele-Operator than traditional large simulator
this information with other vehicles in its fleet. screens. HMD security authentication will be neces-
sary in order to permit secure access to the vehicle in
Tele-Operation order to take over control.
Recent events in 2019 have illustrated the potential Low wireless communications latency is critical
benefits that may be realized if Tele-Operation is used between the vehicle and Tele-Operator and should
to complement AV deployment. One incident involved be at least 10 times faster than the vehicle’s mechan-
a driver sleeping while in Level 2 autonomy mode. ical latency. Therefore, the time for both uploading
There are also complaints from the general public high-quality sensed information, remote processing
that robotaxis can stop unexpectedly and hesitate to and downloading commands will need to be <<100
advance, necessitating a switchover to manual mode. ms. This is particularly challenging for highway speed
In both cases, a Tele-Operator could conceivably take operation and may justify initial deployment with
over control of the vehicle and drive it to safety or robotaxis operating in a city center where vehicle
enable it to proceed forward without holding up traffic. speeds are typically <50 kph (30 mph). Bandwidth
It can be argued, in fact, that Tele-Operation requirements can be reduced with smart data compres-
may be necessary in the near-term for a fleet operator sion protocols and by sending only the “delta” image
because it not only improves passenger satisfaction to the Cloud (and perhaps with only the edges of the
if the vehicle is not overly cautious, but it can also object rather than the entire object).
reduce the manpower to rebalance stranded vehicles. A high bandwidth connection will be needed in
Corner-cases that can confuse AVs and cause hesita- case of decreased performance (multiple users, obstruc-
tion may hurt customer acceptance but tele-operation tions, etc.). It may also be appropriate to select routes
might alleviate the problem and allow a viable robotaxi for where the signal strength is known to be high if it
service to be implemented sooner. Since passengers does not affect vehicle safety or trip duration negatively.
in the robotaxi may need to contact a remote service The vehicle needs to have, at all times, a planned
center operator for other reasons (e.g. for safety, emergency stop function in case connectivity fails. The
security or comfort) this connection may already exist Tele-Operator, ideally, has approximately 10 seconds
between the vehicle and a remote operator. to adjust to seeing a new environment when they are
Tele-Operation solutions require that the: required to engage. Actuator control is relatively straight-
• vehicle senses its environment forward for by-wire vehicles so this should not be an issue.
• vehicle communicates to a remote Tele-Operator Although Wi-Fi can be faster than 3G or even
its 360° awareness of the situation LTE, the latter should be adequate and may enable
• Tele-Operator sends commands to vehicle to large area coverage more affordably. Future 5G
control braking, steering, acceleration systems with pico-cells can further improve Quality
Stereoscopic images or 3D sensor date (e.g. lidar, of Service (coverage, bandwidth, latency). Rather
radar and camera) representing what the vehicle sees is than sharing with other road users, a dedicated
probably necessary because it is more difficult to deter- cellular connection might improve service reliability.
mine depth visualization with monoscopic images. Moreover, each fleet operator may decide to have
Color images are easier for humans to process than its own Tele-Operation center since knowing what
black and white images (shadows, in particular, make causes the Tele-Operation system to engage will be
information for the company and tied in with its be particularly attractive in regions, such as China, where
proprietary software development. homes may not have a charge port and where a remote
The cost of the hardware to enable tele-operation charging infrastructure can be less expensive to install
can be modest, especially for low speed robotaxi opera- and have greater utilization to reduce costs. Another
tion, and it should be possible for one Tele-Operator to example might be driving the passenger(s) in autono-
manage many vehicles since robotaxis typically do not mous mode to a local hospital or caregiver or even, in
fail at the same time and/or can be scheduled to move at some cases, locking the vehicle and driving the vehicle
different times. Robotaxi development can learn from to a local police station. While these corner-cases can
each “remote control” engagement and this should lead be achieved without Tele-Operation, the use of Tele-
to improved algorithms for all the vehicles in the fleet. Operation can make the performance more reliable and
Tele-Operation could become a more active improve the perceived safety of the robotaxi fleets.
area for development because corner-cases can delay
robotaxi deployment and business model monetiza- Changing the competitive dynamic
tion. Some of these applications could include: It is possible that, over time, the AV software becomes
• failure for the driver to respond to take over, as so good that Tele-Operation is no longer required.
requested, in L2+ operation However, in the near-term there are substantial chal-
• taking over if the AV systems degrades below a lenges to solving the many corner-cases and there can be
certain level (either due to system failure or environ- a benefit in leveraging Tele-Operation to accelerate AV
mental conditions) and manual control is impossible commercialization. An AV company wishing to lead in
• augmenting robotaxis in the shared vehicle fleet. market deployment must balance safety, roadworthiness
Other examples might include moving autono- and business viability. How safe is safe enough?
mous, electric vehicles at night to charging stations and One approach to reducing some of this uncer-
returning them to the customer’s parking spot in the tainty might be to leverage connectivity for V2I and
morning prior to being used by the customer; this may Tele-Operation to ensure safety and roadworthiness are
/ŶƚĞůůŝŐĞŶƚsŝďƌĂƟŽŶ/ƐŽůĂƚŽƌƐŽůƵƟŽŶƐ
ĨŽƌƐƵƉĞƌŝŽƌƉĞƌĨŽƌŵĂŶĐĞ
sŝďƌĂƟŽŶŝƐŽůĂƚŽƌƐĚĞƐŝŐŶĞĚƚŽƐƵƉƉŽƌƚĂŶĚƉƌŽƚĞĐƚƐĞŶƐŝƟǀĞ
ĞůĞĐƚƌŽŶŝĐŵŽĚƵůĞƐ͕ƐĞŶƐŽƌƐĂŶĚĐƌŝƟĐĂůĐŽŵƉŽŶĞŶƚƐŝŶ/
ƐLJƐƚĞŵƐĨŽƌůŝŐŚƚǁĞŝŐŚƚĂƉƉůŝĐĂƟŽŶƐ͘
WƌŽƉƌŝĞƚĂƌLJĞůĂƐƚŽŵĞƌďůĞŶĚƐ͕ĐƵƐƚŽŵ
ĨŽƌŵƵůĂƚĞĚĨŽƌƐƉĞĐŝĮĐĞůĂƐƟĐ͕ĚĂŵƉŝŶŐΘ
ĞŶǀŝƌŽŶŵĞŶƚĂůƉƌŽƉĞƌƟĞƐ
WƌŽǀŝĚĞƐĞdžĐĞůůĞŶƚƌĞƐŽŶĂŶƚĐŽŶƚƌŽůŽǀĞƌĂ
ďƌŽĂĚƚĞŵƉĞƌĂƚƵƌĞƌĂŶŐĞ
&ŽƌŵŽƌĞĚĞƚĂŝůƐǀŝƐŝƚŚĞƌďĞƌĂŝƌĐƌĂŌ͘ĐŽŵͬĂƵƚŽŶŽŵŽƵƐ
+HUEHU$LUFUDIW6HUYLFH,QFŇVDOHV#KHUEHUDLUFUDIWFRPŇZZZKHUEHUDLUFUDIWFRPŇ
acceptable from the start and to have the AV software smooth traffic flow and increase throughput for the same
learn from these connectivity interventions. This might average traffic speed, or level of congestion. In addition,
allow an AV company having slightly inferior AV future urban transport systems will need to emphasize
software to still offer the best overall AV performance. multi-modal integration and encourage diversity of
Over time, improved software from machine learning transport. This includes “first-mile/last-mile” micro-
could allow reliance on Tele-Operation connectivity mobility modes, so that the optimal solution for indi-
to be reduced while I2V connectivity should evolve viduals and the city as a whole are more closely aligned.
to V2X as more vehicles are able to communicate It can only effectively be achieved when there is wireless
with the infrastructure and with each other. This is connectivity linking the different modes. ■
likely to happen as vehicles increasingly come with an
embedded cellular connection (to support enterprise
diagnostics and OTA updates as well as to provide Dr. Chris Borroni-Bird is co-author of Reinventing the
infotainment) and C-V2X functionality can be added Automobile: Personal Urban Mobility for the 21st Century, with
relatively easily and affordably. This should reinforce Dr. Larry Burns and the late Prof. Bill Mitchell (MIT Press, 2010).
safety and might even allow AV costs to be reduced He has led advanced automotive-related activites at Chrysler,
over time, as sensing hardware might be simplified GM, Qualcomm MIT Media Lab and Waymo. He is the founder
of Afreecar LLC, where he consults on future mobility and is
if V2X connectivity is ubiquitous in the geo-fenced developing a novel solar-powered solution for the developing
location or even more broadly. world. He holds 50 patents, many related to the ‘skateboard’
The development of safe and well-behaved AVs is platform concept.
essential to their social acceptance and implementation,
but it is insufficient from the perspective of optimizing the For more of Dr. Borroni-Bird’s insights, see March 2018 AVE,
Reinventing the Automobile’s Design [http://www.nxtbook.
road transport system. City planners are also aiming to com/nxtbooks/sae/ave_201803/index.php#/12] and January
improve traffic and connectivity will be needed to ensure 2018 Automotive Engineering, Who Wants AFreecar? [http://
that the actions of independent AVs are co-ordinated to www.nxtbook.com/nxtbooks/sae/18AUTP01/index.php#/28].
Make money from initial broadcasts and heavily Sell driverless vehicles to transportation companies
Value Proposition
monetize archives. and consumers.
Complex data capture and processing involves Develop new algorithms to improve vehicle safety,
Workflow shooting, high-speed ingestion, multiple editing and continually apply new simulation and analysis
processes, and finalizing for distribution. against source data.
Getty Images
Quantum R-series data
storage strategy.
Quantum
Multiple editing teams work on film, sound, data. Advanced storage systems must support equally
special effects, animation and color; they output the advanced security and data protection.
final work by collaborating with each other and with Automated vehicles are a much newer business
directors and producers. This level of collaboration than M&E, but this new “industry” also depends on
and intensive processing requires storage with high generating and working with massive amounts of data.
performance and massive capacity. The automotive sector already generates significant
Once the works are broadcast, M&E continues to data from its AV test vehicles, typically producing 5
monetize them. The rerun is the granddaddy of mone- terabytes (TB) to 30 TB per vehicle per testing day.
tization, but M&E also incorporates sections of existing Researchers build and improve the data intake devices,
works to create new ones. analyze massive incoming data and adjust their algo-
Cyberthreats to M&E also are a major concern: rithms for better analysis and testing.
witness Sony losing millions of dollars to a multi- Long-term data retention and accessibility are
pronged attack in 2014. M&E risks losing monetary crucial to the industry; AV retains data for new research
and reputational value when something happens to its projects and to validate current ones. As a highly
regulated industry, most auto companies plan to keep archive infrastructure provides the data durability and
20 years of test data for liability protection. protection needed while also offering active access to
Unlike M&E, autos are not yet a popular hacker all user data.
target. But data is not only at risk for malware; IT errors, Don’t depend only on cloud-based cold storage.
lost mobile devices, failed hardware and software data Even if you can achieve desired data transport speeds
and even the old-fashioned natural disaster can lose back and forth from the cloud, large-scale storage and
or corrupt data. An industry that depends on original retrieval from cold tiers is slow and expensive. Use
data for its success cannot afford that level of loss. post-production cloud applications or cloud-bursting
to support intensive processing but keep massive data
Best Practices for data storage storage and active archives on-site.
Plan your infrastructure for both today and tomor-
row’s projected data amounts. Buying for current data Build the foundation
is a cheap option in the short-term, but data growth is Data-intensive industries like mobility/AV and M&E
accelerating and will likely outpace the system’s capacity need dynamic scalability without sacrificing availability
and scalability in 1-2 years. Replacing the system will or performance. These industries also need storage
cost even more. with user-friendly, sophisticated management tools and
Plan for a highly-scalable system that will grow smooth integration with workflow and analysis software.
along with big data and user needs. Build your infrastructure from the start to accom-
Don’t depend on disk/flash storage alone. If you modate all the scalability, flexibility, protection and
depend on storing data on your production systems, accessibility you are going to need for the foreseeable
affordability will go out the window. Combine high perfor- future. Do it smart from the beginning and let your
mance disk/SSD production systems with tape libraries for system grow with you. ■
dynamic scalability and highly accessible archives.
Invest in a storage solution with integrated data
Jason Coari (left) is director, Scale-out
protection. AV and M&E need to protect their valuable Storage Solutions at Quantum. He leads
data against intrusion and data loss. Use updated data the company’s product and commercial
protection and remember that tape’s “air-gap defense” strategy for scale-out storage across all
is impassible for would-be hackers—malware cannot industries.
traverse the separation between a system and an
off-line cartridge.
Mark Pastor is director for archive solutions at Quantum. He
For multi-petabyte environments that include represents Quantum within the Active Archive Alliance, the
active archive data, traditional backup apps are not LTO Consortium and the Object Storage Alliance. He regularly
very efficient or effective. A self-protecting active blogs on topics relating to data protection and archival storage.
Radar and lidar get all the attention, but Inertial Measurement
Units are the backbone of sensor fusion. Suppliers are scrambling
to make IMUs more accurate—and much less expensive.
The emerging era of highly-automated driving bring different types of data together from GNSS
comes courtesy of much-vaunted sensor tech- (Global Navigation Satellite System), lidar and cameras
nology—spinning lasers, penetrating radar, sonar to make everything consistent and smooth.”
blips. But the sensor that gets the least respect in Moreover, an IMU can mean the difference
the technology stack could be the lynchpin for the between life and death. The unit’s main components—
mass roll-out of self-driving vehicles: the inertial an accelerometer and gyroscope—determine a vehicle’s
measurement unit (IMU). velocity, acceleration, heading and turn rate. IMUs
“IMUs are the glue that binds everything together,” provide a reliable “ground truth” even when there’s a
said Mike Horton, chief technology officer at Aceinna, gap in GPS, a vehicle heads into a blind corner, or lidar
a Boston-based company that develops sensing solu- struggles with a snowstorm.
tions for automotive applications. “An IMU can help “When other sensors fail, the IMU is the sensor
that gets you to the side of the road to a safe spot,” said
Horton. “It’s uniquely positioned to do that because it
has no external dependency.”
Aceinna
than a century. But wide use in the modern era came
with the development of micro-electro-mechanical
Aceinna’s OpenIMU300 series is a high-performance, open-source inertial plat-
systems (MEMS). Today’s miniature accelerometers
form featuring a three-axis accelerometer, three-axis gyro and magnetometer.
and gyros are soldered to circuit boards and strapped
into vehicles in enclosures about the size of a matchbox.
Basic IMUs are used to detect sudden deceleration to Horton, Aceinna’s CTO, agrees that requirements
deploy airbags and enable stability control. They also for accuracy, size and price are increasing. “I’ve been
orient your smartphone. surprised how aerospace-like the requirements have
IMU sensors for airbag deployments are available been for IMUs in autonomous driving,” he said. “The
for tens of dollars from traditional auto suppliers. At big U.S. automotive companies want really high perfor-
the other end of the spectrum, sophisticated IMUs mance.” At the same time, he said the heart of the auto-
from Honeywell and others are used in high-preci- motive IMU market remains at under $500 per unit.
sion aerospace and tactical military applications. The One of the biggest challenges is to correct for
step up from MEMS to fiber-optic gyroscopes (FOGs) gyro drift, often indicated as “in-run bias stability”
brings greater accuracy, but FOGs cost tens of thou- measured in degrees per second. “No matter what,
sands of dollars. a gyroscope will drift over time in terms of a zero
Critically, IMUs also run software—with some point,” said Scott Kimbrell, senior field applications
providers focusing primarily on algorithms for sensor engineer at Xsens. Engineers also battle the effects of
fusion. Aceinna’s open-platform IMU, for one, allows sensor “noise,” such as angular random walk and bias
end users to run their own algorithms. estimate. These factors could result in a vehicle moving
Unfortunately, the vague nomenclature applied toward an adjacent lane (or worse) when there are gaps
to IMU variants creates confusion. The lines between in input from GNSS or other sensors.
IMUs and inertial navigation systems (INSs) easily get Horton said the goal is to maintain 30-cm (11.8-
blurred, making it difficult to make apples-to-apples in.) accuracy for between a few seconds and a couple
comparisons of specifications and cost. of minutes. “One degree per hour is the sweet spot
“IMU is an umbrella term to describe a wide that we see in autonomous now,” he said. Aceinna’s
assortment of inertial systems,” said Jaya Krushna OpenIMU300 series, which typically offers six-degree-
Panda, an industry analyst with Technavio, a London- per-hour accuracy, sells for $200 or less.
based market-research firm. Technavio pegs the global Xsens’ newest line of IMUs includes the MTi-600, a
automotive IMU sensor market at U.S. $1.13 billion in high-end unit with advanced features and a gyro capable
2018—mostly for rudimentary safety functions rather of 8 degrees per hour; it sells for $300 when purchased
than automated-driving; the number of leading IMU in quantities of at least 500 units. The company says the
providers targeting vehicle-automation applications in-run bias stability for its gyros will continue to improve,
is about a dozen. but an accurate MEMS gyro is less important than the
power of sensor-fusion algorithms to generate highly
Automated driving changes everything accurate readings for pitch, roll and yaw.
“Going forward, our sensors will continue to get smaller Meanwhile, MEMs manufacturing continues
and more accurate,” said Erica Zelazny, sales director to improve. While a one-degree-per-hour IMU is
at Xsens, a provider of inertial technologies for sensor not yet commercially available at automotive-grade
fusion based in the Netherlands with its Americas prices, Aceinna promised a June 2019 introduction
headquarters in Los Angeles. “Of course, everybody for a new IMU offering about two degrees per hour
also wants them to be cheaper,” she added. for less than $100. ■
The sensing technologies needed for auto- technology-assist. Materials-science specialist BASF
mated-driving vehicles are evolving as the industry is enhancing the understanding of what materials may
moves toward high-level (SAE Level 4-5) automa- aid the sensing equation.
tion. Sophisticated sensors already enable advanced “Each step of autonomous driving requires more
driver-assistance systems (ADAS) features such as sophistication and more surety that the sensing data
adaptive cruise control, park assist, lane-centering is correct,” said Mark Minnichelli, director of tech-
and others. nical development for Performance Materials at
But in order to meet the needs of high-level and BASF Corp.
conditional autonomy, sensors themselves need a Signal integrity is mandatory in order for the
sensing signal to be sent and received, according to
Dalia Naamani-Goldman, BASF’s market segment
manager for Transportation Performance Materials.
“On the radar side, we have transparent materials and
are developing an absorptive material that can be used
on the sensor surface, housing, and mounts,” Naamani-
Goldman said, noting the material is being tuned to
function between 70-90 GHz.
A non-variable production process is crucial.
“There is speculation in the industry that the frequency
requirement could go up to 150 GHz. So if there is a
shift in the frequency requirement that the material
must function in, we’ll need to look at different additive
packages, as well as determine what filler content will be
added to the virgin polymer,” said Naamani-Goldman.
Sensing clarity is another autonomous driving
BASF
Current dark exterior vehicle paint colors can impede lidar signal transmis-
must-have.
sion. BASF technology specialists are developing near-infrared reflective Whether the sensor is radar, lidar, or even a
dark automotive exterior paints that are lidar-compatible. yet-to-be-developed technology, the autonomous
German University
Studying Assisted-
Driving ‘Handoff’
by Stuart Birch
AB Dynamics’ handoff
simulator has the ability
to change suspension
parameters on-the-fly
or to conduct complex
automated maneuvers
in chaotic traffic
conditions.
TS-7553-V2
under a variety of circumstances.” Industrial IoT Gateway
The aVDS has been designed to deliver the university’s NXP i.MX6UL 698MHz ARM Cortex-A7 CPU
required level of driver immersion, providing the necessary
4 GB MLC eMMC Flash
dynamic ability together with innovative visual, audio and
haptic assets, and vestibular cueing. Schick stressed the
significance of high-dynamic capability and low-latency
response times.
The university also plans to use the simulator’s archi-
tecture for the mounting of a real car steering system on
the platform. It will be complete from steering wheel to ball
joints. The reason, said Schick, is a steering rack can be diffi-
cult to model because it has many non-linear components.
There are significant advantages to its inclusion within the
simulator’s feedback loop, avoiding any inherent inaccuracy
of a modeled system. Steering-based testing will include
evaluation of on-center feel and driver response.
AB Dynamics Europe Managing Director, Klaus
Weimart, said: “ADAS and autonomous systems will play
a big part in future mobility solutions and the way we as Nine-Axis MEMS MotionTracking Device
humans interact with them is paramount to their success.”
Industrial Temp Range -40 °C to 85 °C
The company’s aVDS incorporates software from rFpro.
by Bill Visnic
As the product-development landscape for Perhaps the most dramatic demonstration comes
light-vehicle electrification and automated-driving from ZF’s Vision Zero concept vehicle, a rolling
technologies becomes less cluttered, it’s apparent that showcase of in-development technologies combined
established automotive Tier 1 mega-suppliers are with current-production components to express the
intent on merging their established competencies with “what if ” goal of the Vision Zero: complete elimination
whatever new product lines are required in the electri- of accidents and vehicle-borne emissions.
fied, automated future...whenever it comes. The Vision Zero concept rolls on ZF’s produc-
This strategy was recently on display at a tech- tion modular semi-trailing-arm rear suspension that
nology background event conducted in the U.S. by also integrates an electric-drive axle and rear-steering
ZF. Long regarded for its transmission, driveline and capability. It makes for a decidedly complete rear axle
chassis expertise, ZF showed customers and jour- that could theoretically be “added” to virtually any
nalists various prototype systems intended to merge light-vehicle design, ZF’s engineers expounded.
current-generation components to create higher-level But the concept car also demonstrates ideas for
functionality—and value. next-generation vehicles equipped with advanced
driver-assistance systems (ADAS) functionalities.
Zero accidents and emissions There’s an intriguing Driver Distraction Assist system
At the event’s test track, ZF demonstrated an advanced that monitors the driver’s state of attention—but ZF is
near-term driver-aid that dovetails with automat- going further with a new Interior Observation System
ed-driving features: Wrong-way Inhibit employs (IOS) that uses a 3D camera to augment information
known road-mapping data to prevent the driver (or from other sensors to determine if the driver is atten-
autonomous system) from entering a one-way street tive and prepared to take back control of the vehicle
or highway in the opposite direction of traffic flow. if it is in an automated-driving mode.
On a small spur of the track that engineers had The IOS also can ascertain the position and size of
designated as having opposing-direction traffic, the other occupants—information that can be leveraged for a
test vehicle automatically came to a halt and refused variety of complimentary safety and convenience features.
further progress in that direction. Due to be production-ready by late 2021, IOS’
“visual data from the interior of the vehicle can be highly
valuable from many standpoints,” said Norbert Kagerer,
senior VP, engineering, for ZF’s occupant safety systems
business. “The development of our 3D interior observa-
tion system leads to enhanced safety and convenience and
helps support the evolution toward automated vehicles.”
To help develop IOS’ full potential, ZF in May 2019
acquired a 90% interest in Germany-based Simi Reality
Motion Systems. The 35-employee company brings deep
experience in image-based 3D systems for the recording
and analysis of human movement. ZF believes such
ZF
Part of ZF’s “Vision Zero” technology suite, the Wrong-way Inhibit system competencies are fundamental to achieving the best infor-
prevents driving into oncoming traffic. mation about the passenger-compartment environment. ■
by Stuart Birch
5G Cellular-Enabled
Test Infrastructure
for AVs Launched at
Millbrook
by Stuart Birch
hyper-dense small cell network that delivers ultra- reference vehicles that can be programmed to emulate
high capacities which enables a range of new CAV use a range of vehicles for AV software development, and
cases to be explored,” said Paul Senior, chief strategy “soft targets” for real-world AV requirements. It also
officer at Airspan Networks. can provide a digital model of the facility’s test tracks
The Millbrook mobile network facility comprises with 1 mm surface accuracy for vehicle simulation, as
77 small-cell base stations (using 2.3 GHz and 3.7 GHz well as a simulator suite for both physical and virtual
spectrum) and 22 millimeter-wave access points around development.
Millbrook’s banked, high-speed road circuit. (Millimeter- With cybersecurity a continuing concern for AVs,
wave is the spectrum from 30 GHz to 300 GHz.) There the 5G testbed can be used to simulate mobile-net-
are 59 masts, 38 for the small cell base stations and 11 for work disruption and connectivity to test software,
the millimeter-wave access points, plus 10 for the milli- vehicle and infrastructure resilience to interference.
meter-wave mesh backhaul along a hill route. Some 19 And an open-access network, with slicing to enable
km (11.8 miles) of fiber-optic cable is installed. users to work on a single network infrastructure, facil-
itates testing involving various network disruption
160-mph coverage and connectivity situations.
The site’s new communications systems provide up To g e t h e r w i t h A i r s p a n Ne t w o r k s ,
to 1 Gbps real-time connectivity; up to 256 km/h the AutoAir consortium includes McLaren Applied
(160 mph) continuous coverage capability around Technologies; Blu Wireless Technology; The 5G
the high-speed bowl; access using 4G long-term Innovation Center, University of Surrey; Dense Air;
evaluation—the network capable of supporting 5G Quortus and Celestia Technologies Group.
New Radio (NR) when commercial devices become Another proving grounds, the Swedish test
available—and secure Multi-access Edge Computing facility AstaZero, now is accredited by Euro NCAP to
(MEC) with the assurance that data does not leave perform tests of active safety systems and automated
Millbrook’s environs. driving. It is owned by Research Institutes of Sweden,
The Millbrook facility offers access to open-source (RISE) and Chalmers University of Technology. ■
Delivering on
Autonomous Transport
by Ryan Gehm
Einride also is collaborating with Swedish logis- an e-motor with direct drive coupling to traditional axles.
tics company Svenska Retursystem to initiate a pilot With a battery capacity of 200 kW-h, Einride estimates
project with the goal of introducing multiple T-pods around a 200-km (124-mile) range on one charge. The
at a logistics center in Västerås, Sweden. electronically limited top speed is 85 km/h (53 mph).
And Michelin announced at the Movin’ On The company has also developed an “up-pow-
Summit in June that it would deploy Einride’s T-pod to ered adaptation” of the T-pod for the logging industry.
transport goods at a production facility in Clermont- Equipped with a 300-kW-h battery, the ‘T-log’ can
Ferrand, France, beginning in 2020. A second phase, carry a 16-tonne load and navigate forest roads.
requiring approvals from the French authorities, will Capable of SAE Level 4 self-driving, the truck’s
involve transporting goods on public roads between operation is supervised and can be operated by remote
different Michelin sites in the city. control when necessary.
Einride has its sights set on the United States “Connectivity is key for Einride,” chief technology
as well. The company has several customers based officer Pär Degerman told AVE. “We can use 4G for
in the U.S., a spokesman told Autonomous Vehicle the installations we have today, but of course 5G will
Engineering, and it plans to roll out commercial instal- become critical over time when we start managing
lations there in the not-too-distant future. fleets of autonomous, electric vehicles.”
Einride has partnered with Ericsson to explore
360-degree awareness the opportunities for 5G mobile connectivity. The
Because the T-pod has no driver’s cab, the vehicle can two companies demonstrated the technology earlier
be smaller—approximately 7 x 2.5 m (23 x 8.2 ft)—but this year at the Mobile World Congress in Barcelona,
with increased loading capacity compared to a conven- allowing show-goers to remotely operate a T-pod.
tional truck. It can transport 15 euro-pallets and weighs “Our main concern now is that operators start
26 tons fully loaded, the company claims. rolling out 5G as soon as possible and that they rethink
The vehicle’s drivetrain concept is centered around how they build their mobile networks,” Degerman said.
Einride
Equipped with a 300-kWh battery, the T-log is an “up-powered adaptation” of the T-pod that can carry a 16-tonne load.
Driverless Ridesharing
Remains in Uber’s Sights
by Kami Buchholz
Uber hasn’t set a timeline for deploying auton- road tests that can prove out things that we’ve already
omous vehicles (AVs) on its worldwide ridesharing checked in simulation and on the track,” he said in an
network, but plenty has happened since the company interview with SAE’s Autonomous Vehicle Engineering.
began development work in 2015—the latest being More than 1,000 engineers and other specialists
the June 2019 introduction with close partner Volvo are working at Uber ATG hubs in Pittsburgh, San
Cars of a self-driving vehicle the companies claim is Francisco and Detroit, as well as in Toronto for R&D
production-ready. projects. Volvo’s XC90 SUV—the same platform on
“The base vehicle is equipped with key safety which the new self-driving model is based—has been
features that allow Uber to easily install its own self- Uber’s autonomous-driving test fleet mainstay. Other
driving system, enabling the possible future deploy- vehicles are slated to join the development program.
ment of self-driving cars in Uber’s network as an auton- Uber and Toyota announced a deeper AV collabo-
omous ridesharing service,” Volvo announced. ration in August 2018. Each company’s technologies will
The company stated that when paired with Volvo’s
vehicle platform, Uber’s self-driving system “may one
day allow for safe, reliable autonomous ridesharing
without the need for a ‘Mission Specialist’”—Uber’s
term for its employees who are trained to serve as
safety drivers while the car operates in areas designated
and suitable for autonomous drive.
In December 2018, nine months after one of its
self-driving Volvo development vehicles was involved
in a pedestrian fatality in Tempe, Arizona, Uber
restarted AV test drives on select public roads.
“A larger and larger amount of our testing is
Volvo Cars
Kami Buchholz
a hardware kit,” Lesh said.
Field-of-vision studies for sensors using CAE
analysis is a prime point of investigation. “Compared
to designing human-driver sightlines via the seating
be integrated into purpose-built Toyota vehicles with the position and pillar locations, we need autonomous ride-
initial Autono-MaaS (autonomous mobility as a service) share vehicles to be set up with all the sensors having
fleet based on the Toyota Sienna minivan platform. an optimum field-of-view to the surroundings,” Lesh
According to a Toyota Research Institute spokesperson, said. “That’s why we’re doing a lot of work relating to
vehicle deliveries are slated for 2021. Said Lesh, “We plan the lines of sight.”
to work with multiple OEMs now and in the future. Uber drivers around the world complete 15
Although we’ve announced some collaborations and million rideshare trips daily. “We want autonomous
partnerships, we haven’t announced everything.” vehicles to supplement that network, not replace it.,”
Lesh said. Today, there aren’t any Uber network trips
Technology-acquisition options with AVs. “It’s hard to predict timing, but regulatory
Uber’s test vehicles are retrofitted with DC-DC actions need to occur and the public needs to trust
converters and other off-the-shelf hardware, but not all of that the autonomy is verified and that it’s safe. Those
the needed technologies are commercially available—or factors will drive our network rideshare deployment,”
available from a supplier. “If a technology doesn’t exist, Lesh maintained. ■
Hosted by: