Mitsubishi Electric is investing in the development of a high-accuracy 3D map of Japan, aimed at all stakeholders in the automotive industry. Image supplied.
This feature was first published in issue 99 of Position magazine. Stay up to date with the latest research and industry trends by subscribing here.
It’s a chimeric promise of Utopian proportion or dire dystopian cast, depending on your relationship to your vehicles.
Seamless, personalised road transport that you barely need to interact with. Your car an extension of your living room, commutes an expansion of leisure time or productive potential. Safe, silent, automatic vehicles interacting with their surrounds in the most efficient, stress-free manner possible. Bid farewell to road rage, frustration at the wheel, gridlocked cities and the high likelihood of a violent death — the cars of the future won’t even have a driver’s seat, let alone a dashboard or gearstick.
With over 1,000 road deaths in Australia each year and 94 percent of all traffic incidents attributed to driver error, automated road transport seems to promise a radically safer future on the road, a point agreed upon by public health research and futurists alike. Cityscapes transformed by liberating the roughly 30 percent of the urban environment currently locked up for parking space, freed for humans by the more efficient spatial needs of a smaller, intelligent transport fleet that does not need to remain proximate to its passengers at all times.
Governments on the bandwagon
Back to reality, you say? Mirage or not, know this: Australian governments are getting ready. The Australian Communications and Media Authority (ACMA) issued a Radiocommunications (Intelligent Transport Systems) Class Licence in late 2017, allowing connected vehicles and smart road infrastructure to share data on the 5.9 GHz radio frequency band, a certification aligning with international standards for nations serious about a driverless automotive future.
2018 saw a wave of initiatives and attendant investment commitments across the country, as trials were extended, pilots announced — typically partnerships with a specific manufacturer to test an autonomous vehicle or smart piece of infrastructure in carefully controlled conditions or spaces.
Western Australia’s RAC announced a significant expansion of its automated vehicle program with the purchase of a $490,000 NAVYA Intellicar prototype, with designs for this to become a fleet to be trialled on public roads in the first half of this year, an expansion of its testing of the Intellibus. Bosch was the first manufacturer to receive a Victorian permit for automated vehicle tests on public roads in January, receiving a $2.3 million grant to test its developing driverless tech on high speed, rural public roads. Their autonomous platform is developed and manufactured by Bosch Australia, and based around a Tesla Model S. But perhaps the most significant milestone on the road towards a driverless reality was the announcement of a common legal framework to support automated vehicles, which made national headlines in May 2018.
The RAC’s plan to build a fleet of NAVYA Intellicars will build on its low-speed Intellibus trial, which has ferried over 14,000 people around the South Perth foreshore.
Hard reset: who’s at the wheel?
Initial reporting of this announcement was fairly light on detail around the content of such a national law, but its status as a turning point in the road automation agenda is significant. A patchwork of over 700 existing state and federal laws governing road use would need to be updated to support safe and accountable activity of autonomous vehicles, with the existing legal framework simply inadequate to cope with the complexities of a realm in which responsibility is not wholly attributable to a human driver. A national framework demonstrates a commitment to a swift, blanket solution to fast-tracking this process.
Australia’s National Transport Commission (NTC) has been doing much of the hard yards in the push to prepare Australia’s regulatory environment for this new reality. Commencing on a program commissioned by the Transport and Infrastructure Council in 2015, they have been researching, consulting and pushing out policy papers in pursuing their ‘roadmap to reform’ agenda. This initiative aims to achieve a safe operating environment for conditionally automated vehicles on Australian roads ‘before 2020’, and for highly and fully automated vehicles by 2020.
The NTC established the guidelines under which automated vehicle trials are operating, and in a 2017-18 review of Australian driving laws, found that the core of many of these legal issues is that Australian transport law assumes a human driver. Areas such as injury insurance for those involved in an accident with an automated vehicle, regulations around human supervisors for automated driving systems (ADS), vehicle compliance and enforcement methods were all found to need revision to create a legal environment that accounts for the risks and intricacies of this new legal reality.
One law to bind them all
The proposed common framework is a one-size-fits-all approach designed to address the raft of glaring issues in one fell swoop — a national law that would override and interface with all state discrepancies to provide a universally applicable legal environment that manufacturers and OEMs can reckon with. Indeed, it’s an approach they have been lobbying for internationally, and was the first recommendation in the NTC’s policy paper on driving laws — and now it’s an approach that all state and territory transport ministers have agreed to.
Dr. Kirsten McKillop, manager of automated vehicles at the NTC, said that stakeholder feedback overwhelmingly supported a purpose-built national law as the most efficient approach to a fit-for-purpose framework, due to the extent of the changes required.
“We’ve had strong feedback during our consultation to not try and tinker with individual laws, although there will be a need for some consequential amendments etc. — that we should be approaching it as something entirely different that needs to be regulated to ensure safety. It is quite a different idea than a human driver,” she said.
“Most of the compliance for drivers at the moment is targeted at individual drivers, whereas with automated vehicles you’re looking more at corporations. So that makes a difference to what type of compliance mechanisms might be useful.”
An issue at the heart of a reform package for driving law is the distribution of responsibility for the act of ‘driving’ across a range of entities, which could now include corporations, proprietary software systems that manage a suite of systems to operate the vehicle and interface with humans to varying degrees, and humans that share responsibility for control of the vehicle under certain circumstances.
Following a fatal accident in March 2018, in which an Uber self-driving SUV killed a pedestrian whilst undergoing public road testing in Arizona, Australian state transport ministers agreed that regulation must stipulate conditions governing the fitness of human supervisors for their involvement in an automated driving system. Uber’s system could be categorised as within the controversial level three automation, in which a human observer is essentially the last line of defence, or fallback user, for avoiding obstacles or environmental challenges.
Stay up to date by getting stories like this delivered to your mailbox.
Sign up to receive our free weekly Spatial Source newsletter.
“We are looking at a purpose-built law for automated vehicles, because the tasks and responsibilities are in some ways quite different to the current rules that apply to humans,” Dr. McKillop said.
“Level three [automation] for example, with a fallback-ready user — that’s a new party who hasn’t previously been regulated. Ministers decided in May 2018 that that fallback-ready user should be regulated so that they should have some duties to be sufficiently vigilant to take back control — be appropriately licensed, unintoxicated and so on.”
Dr. McKillop said that the NTC’s five overarching recommendations for the content of such a law are that it would firstly allow an automated driving system that has been approved under and continues to comply with the safety assurance system to perform the driving task. Secondly, a national law must ensure there is always a responsible legal entity for the driving task, thirdly clarify who that responsible entity is, fourthly set out any obligations on relevant entities, including the automated driving system entity, and users of automated vehicles and finally: to provide a regulatory framework with flexible compliance and enforcement options.
If it sounds like a tall order, that’s because it is. The NTC is essentially driving the reform agenda to develop this law, presenting recommendations to the transport ministers for the states at bi-annual meetings of the Transport Infrastructure Council. At the next, to be held in May, the NTC will make recommendations on changes to insurance schemes to ensure that anyone involved in an accident with an autonomous vehicle is not worse off relative to the victim of a conventional road accident, along with recommendations on privacy, surveillance and government access to data.
Sensor fusion: some of the common sensors utilised in Advanced Driving Assistant Systems (ADAS) and Automated Driving Sytems (ADS). Image: Texas Instruments.
The bleeding edge: safely certifying new tech
There’s more to be considered than the piece-by-piece dissolution of the human driver in the eyes of the law, and the cyberpunk ethical questions it conjures up.
Highly autonomous road vehicles are extraordinarily complex and rapidly developing systems, composed of many sophisticated subsystems. Approving such products for use on public roads requires a comprehensive and rigorous overhaul of engineering and safety certification procedures.
The NTC is closely watching international developments, such as UN working parties that are considering Automated Vehicles (including changes to international conventions), seeking to align regulatory approaches wherever possible — a wise and necessary move as Australia is likely to be a secondary market to the US and Europe for most manufacturers.
In November 2018, state transport ministers voted to include safety criteria within existing legislative mechanisms, such as the Road Vehicles Standards Act and the Australian Design Rules.Dr. McKillop said that manufacturers would need to self-certify against these criteria to ensure their systems can perform the driving task safely — allowing for technological agnosticism.
It’s important that the certification process remains non-specific on technical requirements, because the road to a driverless future is a dynamic and competitive one. Whilst most systems currently under development rely on the same base technologies — all with their own strengths, limitations and paths to maturity — their roles in the greater automated driving system vary, and time will tell which configuration and application of these capabilities will dominate and stabilise into an industry-wide best practice, if any.
Baraja’s LiDAR scanner design eliminates moving parts for longevity in the demanding road transport environment. Image supplied.
Nick Lansdale-Smith, vice president of business development at LiDAR manufacturer Baraja, is highly aware of this. His firm has developed a novel scanner design, aimed squarely at the autonomous road vehicle market that eliminates mechanised and moving parts within the unit by utilising ‘prism-like’ optics.
“Regulation always follows innovation. Autonomous vehicles are still very much a fledgling technology and will see continuing, rapid innovation for years to come. Regulation today should seek to nurture and support the development of self-driving vehicles safely, cognisant of the fact that the industry still has a long way to go,” he said.
“The biggest challenges relate to the fact that no one sensor modality is sufficient to crack the problem of how to build a safe, reliable autonomous vehicle. Finding the right combination of capabilities and interdependencies of these sensors is a significant challenge facing developers today.”
Secret recipes: a cocktail of technologies
An array of scanners keep the driving system aware of the vehicle’s position relative to other vehicles, lane markings, obstacles and navigational waypoints. On top of the array of electronic driving and navigational aids now standard on contemporary vehicles, vehicles on the road to driverless control are bristling with sensors: LiDAR, optical cameras, long- and short-range radar, ultrasound. Proprietary control systems manage and integrate all of these subsystems to keep the vehicle aware of and responsive to its surroundings.
This bristling suite of sensors provides relative positioning duties, which complement absolute positioning systems — high accuracy GNSS and inertial sensors — to allow an automated driving system to handle the myriad navigational, positioning and manoeuvring duties required in competent operation of a vehicle on public roads.
u-blox are developing extremely high-accuracy absolute positioning schemes for integration in automated driving systems. Image supplied.
Rod Bryant, senior director of technology in u-blox’s positioning product centre, says that at this stage of development, the approach most manufacturers (OEMs) are taking is to leverage the redundancy of this wealth of inputs, using the input of these systems to cross-check each other, and hence distribute risk across the subsystems.
“If the sensors were more closely integrated it would be far more difficult to assign risk to individual sensors. The penalty for this independence is reduced performance and therefore, we should expect an evolution towards tighter integration eventually,” he said.
u-blox has developed a lane positioning system that works with a combination of high accuracy, dual band GNSS, an inertial measurement unit (IMU) and wheel sensors. Bryant said that closer integration of GNSS-based positioning with the landmark-based positioning systems should result in improved ultimate positioning performance by reducing effects of inertial drifts when combined with motion sensing, and calibrating out map registration issues.
Holes in the map
Mitsubishi Electric, a manufacturer of LiDAR systems for automated driving applications, has a similar view of an absolute positioning scheme being central to an autonomous future on the road.
“Positioning correction based on landmark coordinates complements GNSS signals best when those signals are blocked, such as when in tunnels or below high buildings. Camera images and algorithms to acknowledge the target landmark with defined coordinates from those camera images are the key technology here,” a spokesperson said.
“The two are not complementary, they are mandatory. They are both to be used in combination to obtain the basic positioning data, in order to create a high definition 3D map.”
A wicked problem standing in the way of an autonomous driving future are the lack of sufficiently detailed road maps to allow for autonomous navigation with highly accurate features, and the almost incalculable volume of work and investment required to create and maintain such a resource.
Localised initiatives are taking off to cater for specific markets — such as the Dynamic Map Platform Co., Ltd, a commercial entity formed to create a common high precision, 3D base map for use by all entities in the autonomous vehicle space in the Japanese market. DMP was established by Mitsubishi Electric, and designed for use by all automotive manufacturers.
As advanced as this technology is and despite the blistering pace of its development, we are still some way off of seeing driverless production cars operating at scale on Australian roads. A raft of regulatory challenges remain to be resolved in spite of the NTC’s ambitious agenda. Whilst some level 4 trials are kicking off under strict operating conditions, the advanced driving assistant systems (ADAS) of high-end production vehicles currently entering the market — present in new models from Audi, Mercedes — Tesla and Volvo, qualify as level two automation. OEMS working on level 3 and 4 concepts are hard at development, with no plans for production vehicles at this time.
The concept has not been subject to the court of wider Australian public opinion in any significant detail to date. Besides a handful of alarmist articles and some deeper discussion of the technological and ethical considerations from The Conversation and technology-focused outlets, discussion of the issues around it as a genuine reality have been missing from the mainstream. Polling in the US over the past few years has shown that typically over half respondents feel deeply uncomfortable with the idea of sharing the road with, or riding in autonomous vehicles, a sentiment exacerbated by unscrupulous behaviour of tech giants such as Uber and Tesla.
Despite this, once we have a legal framework in place, there will be no further barriers to the integration of increasingly intelligent driving systems infiltrating production vehicles as rapidly as the pace of development allows.
Stay up to date by getting stories like this delivered to your mailbox.
Sign up to receive our free weekly Spatial Source newsletter.
