Misconceptions of autonomous cars

Self-driving cars are a rapidly evolving technology which only a few years ago was still considered science fiction. In such a dynamic context, quick intuitions can be very misleading and misconceptions about the technology, its impact, and the nature of the innovation process abound. In a short article we examine the following four misconceptions:

  1. Driver assistance systems will evolve gradually into fully autonomous cars
  2. The first models of fully autonomous cars will be targeted to the consumer and will be available for purchase
  3. It will take decades until most of the vehicles on the road are capable of autonomous driving
  4. Self-driving cars are controlled by classical computer algorithms (if-then rules)

Driverless car revolution: Buy mobility – not metal

driverless-car-revolutionA new book by Rutt Bridges examines the impact of autonomous vehicle technology on mobility. It is an excellent read, a thought-provoking book which paints a very detailed picture of the future of mobility. It is a wake-up call for the auto industry and a must-read for anyone involved with transportation policy.

Book description:

Imagine a future without congestion, car crashes, smog, or road rage. It’s coming sooner than you think. Summoned with an Uber-like smartphone app, driverless cars will revolutionize transportation. For less than bus fare you’ll enjoy the quiet, comfortable door-to-door service you’d get from a personal chauffeur. A chauffeur that is never distracted, never tired or testy, and always knows the fastest and safest route to get you where you’re going. No cash, no tipping, no crowds, no congestion – just hop in, enjoy the ride, hop out, and be on your way. These cars will be electric: quiet, clean, and so safe that deaths and disabilities will be rare. Instead of dealing with road rage and the frustration of bumper-to-bumper traffic, you’ll be free to text, Facebook with friends, or get a head start on your workday. Since you can cut your cost in half by riding with another passenger, seamlessly arranged by your mobility provider, traffic congestion will slowly fade away.
Owning a car means car payments, insurance, registration, maintenance, gas prices, smog, tickets, accidents, finding parking, and dealing with the stress of traffic. Buying miles instead of metal means you’ll save thousands a year for your dream vacation, the kids’ college education, or buying a home of your own. In addition to lowering stress and regaining the use of 5% of your waking hours, putting an extra $5,000 a year in people’s pockets will compel this change.
Driverless Car Revolution explains the benefits for people of all ages, from kids through seniors, plus the disabled, the working poor, tourists and other special groups. The book also discusses the economic disruption of major industries as well as potential geopolitical upheavals – all the pieces of the puzzle, and how they fit together.
Fasten your seatbelt, engage, and prepare to join the Driverless Car Revolution.

Get it via Amazon.com

This graphic shows the future of the auto industry

It may have taken professional auto industry analysts some time to understand the impact of autonomous vehicles. But now Morgan Stanley’s Adam Jonas has come up with another ingenious 2-by-2 chart so much en vogue with international strategy consultants which highlights the core transformative forces at work. This is a major intellectual feat because it compresses the problem space and helps reason about changes, challenges and opportunities associated with self-driving cars. The chart is shown below. Because I don’t have access to the original Morgan Stanley report the following explanations may not exactly reflect Morgan Stanley’s reasoning.

End-of-the-auto-industry

Quadrant (1) shows the auto industry today which is exposed to to major forces of change: The sharing economy leads to the emergence of companies which provide mobility as a service. Uber, Car2Go, DriveNow, Lyft and others are examples for this trend. In parallel, the auto industry faces the trend toward autonomous driving. Several companies, including Daimler, Nissan, and others are working on models targeted toward the consumer which can drive autonomously. The fourth quadrant shows the confluence of both trends: The shared autonomy. Autonomous pods such as the Google electric autonomous 2-seater, the Lutz Pathfinder currently being deployed in the UK and CityMobil2 autonomous buses fall into this category.

The future of the auto industry can be found in this fourth quadrant. Economic reasons clearly show that this quadrant will capture the lion’s share of individual motorized mobility. Neither of the other quadrants will be able provide individual mobility at competitive prices compared with the providers of autonomous mobility services of quadrant (4). Of course the other quadrants – particularly the 1st and third quadrant will not disappear entirely. There will still be some privately owned cars but they will represent a much smaller share of the mobility market than today.

Adam Jonas conclusions about the future of the auto industry are in line with the scenarios I have outlined over the past years in many articles – including five years ago in my first paper on this topic:  ‘Autonomous cars: The next revolution looms‘ .

Source: Morgan Stanley, Los Angeles Times

Autonomous long distance trains moving forward

Being fixed to a track, trains are much better suited for autonomous operation than road-based vehicles. But  most of the innovation in autonomous vehicles is occurring on the road. Worldwide there are only very few efforts to develop autonomous trains (automated subways and metro lines are not autonomous – their cars are usually controlled by a central server and these lines require significant extensions of the track-side sensors and safety mechanisms which doesn’t scale for long distance rail networks). Fortunately some isolated efforts are now moving autonomous trains forward:

Global mining powerhouse Rio Tinto operates its own 1700km rail network in Australia to transport iron ire from its 15 mines to the sea ports. The company is spending more than 500 million USD to equip all its locomotives with radar, sensors and mapping technology for autonomous operation. The first trial runs have been completed successfully at the end of 2014 and up to 41 autonomous trains may begin operation in the second half of 2015! Is it a surprise that these autonomous trains are being developed by a commercial company that has its own extensive rail network rather than a traditional railway operator?

Although autonomous trains could significantly lower costs, increase capacity and flexibility, most railways are heavily regulated and are unlikely to adopt autonomous driving technology on long distance trains soon. This is unfortunate because the extreme focus on safety actually prevents useful innovations from being adopted and pushes people to other transportation mediums such as the road – with much higher risks and casualty levels.

Fortunately, the effort to develop a European Railway Traffic Managment System (ERTMS) has laid some groundwork which could be leveraged for autonomous operation: ERTMS distinguishes four levels of train control: Levels 0 to 2 rely on standard trackside infrastructure for train control – including signs and balises (transponders embedded in the track which digitally transmit location and track constraint information to the train ). But level 3 allows trains to localize themselves via sensor and retrieve track constrains and movement authority via mobile internet (GSM-Rail). This greatly increases flexibility and should simplify the introduction of autonomous railways on the many routes that are not yet equipped with automated train control infrastructure.

Netherlands first to operate a self-driving shuttle in public traffic?

The competition for low-speed self-driving vehicles in public traffic is heating up. Now the executive council of Dutch ministers has given the green light for running two driverless shuttles in the Dutch city of Wageningen starting in December 2015. The electric shuttles will carry up to 8 persons from a train station to the university on a stretch of approximately 6km on public roads with a maximum speed of 50km/h. Although these will be tests, the shuttles will operate autonomously without safety drivers on board. The shuttles’ operations will be monitored remotely. Before the shuttles be placed in service both chambers of the Dutch parliament need to amend Dutch traffic law. If everything goes according to plan, the world’s first fully autonomous shuttles without backup driver on board could make history in the Netherlands in December!

© Ligier Group

Image: EZ-10 Autonomous Shuttle of Ligier Group, Easymile

Sources: de Gelderlander, carrepublic.nl

Accident rates of self-driving cars: A critique of the Sivak/Schoettle study

To what degree are self-driving cars likely to reduce accidents and traffic deaths? This is a very important but very hard question which has implications for testing, insurance, regulations and governments considering to accelerate or delay the introduction of autonomous cars. Now two researchers, Michael Sivak and Brandon Schoettle, of the Transportation Research Institute at the University of Michigan have examined this problem in a short study titled “Road safety with self-driving vehicles: General limitations and road sharing with conventional vehicles and arrived at four conclusions which – when read carefully – provide little insight into the problem but when read casually seem to raise doubts about the expectation that self-driving cars will be significantly safer than human drivers.

As an example the abstract summarizes their second conclusion as follows: “It is not a foregone conclusion that a self-driving vehicle would ever perform more safely than an experienced, middle-aged driver”.

Who could argue against this statement? Of course, this is not a foregone conclusion. This is a hard problem and a substantial question. Neither would it be a a foregone conclusion that a self-driving vehicle would ever perform more safely than an experienced, young driver (or even an unexperienced young driver). But many readers will interpret this conclusion that the authors – after having analyzed the issue – have found substantial problems that raise doubts as to whether autonomous cars could ever perform better than experienced, middle-aged drivers. But the full text of the report contains just one sentence which further examines this problem:

“To the extent that not all predictive knowledge gained through experience could exhaustively be programmed into a computer (or even quantified), it is not clear a priory (italics by the original authors) whether computational speed, constant vigilance, and lack of distractability of self-driving vehicles would trump the predictive experience of middle-aged drivers”. (Page 4)

Nobody can argue with this statement. It would be a good introduction to a chapter that looks at this problem in more detail, provides some framework, examines the different aspects etc. etc. But this does not materialize.

If we read the study carefully, then we find a pattern that valid questions are being raised, a small number of the aspects relating to these questions are outlined, and then the questions are rephrased into conclusions which themselves are questions. This is unfortunate because the topic is extremely important. More than a million people die in traffic accidents every year. If – twenty years from now – we might look back from a situation where traffic accidents have fallen by more than a factor of five, then we will be able to state with certainty how many lives could have been saved if self-driving cars would have been introduced a few years earlier. We might find that tens of thousands of people have lost their lives because governments and regulators did not realize the risk of delaying a highly beneficial technology and business and innovators were reluctant to advance the technology because of a climate of mistrust and skepticism with respect to the technology. Of course, from the perspective of today this is not a foregone conclusion but we need to make an effort to understand the risks and likely accident patterns of autonomous vehicles much better.

There are lives at stake both if we are too optimistic and too pessimistic over the potential of this technology. But the problem is not symmetric: If we are too pessimistic with respect to the potential of this technology, then we can easily find ourselves in a situation in the future where we find in hindsight that thousands of lives have been lost because of this pessimism and the resulting delay of the introduction. On the other hand, if we are overly optimistic with regard to the technology, and accelerate innovation in this area, it is unlikely that thousands of lives will be lost because the cars do not perform as safely as expected. We can be confident that certification bodies will do their work and uncover problems before they can cause thousands of deaths and regulators will most surely step in immediately when these cars do not perform as expected. At the current stage therefore, pessimism about the technology’s potential may be much more deadly than optimism (which should not be confounded with being blind about the risks).

We should work together urgently to formulate a theory of human traffic accidents and self-driving car accidents which can help us shed light on the issue and understand and organize the many different aspects of this problem. This is hard but it can be done. Please contact me at info.2011 ( at ) inventivio ( dot ) com if you are already working on this topic, if you know of a suitable approach for covering this problem or if you are interested in working together on this topic. I will post one approach on how this could be achieved next week.

Changes:
2015-01-23: Added link to the full text of the study.

 

Global technical regulations for autonomous vehicles: Informal working group established

As regulators grapple with autonomous technology, conflicts between country-specific laws could impede the adoption of this technology. The United Nations has a forum (“WP29“) which aims to avoid such problems by harmonizing vehicle regulations. Many aspects of technical regulations for wheeled vehicles are discussed in a broad range of (informal) working groups. Because of the rapid progress of autonomous technology, the informal working group on Intelligent Transport Systems has recently been renamed and refocused as informal working group on ITS/Automated Driving.

The participants are now laying the ground work for future regulations. They have discussed various approaches to frame levels of autonomy and seem to be leaning toward SAE’s 6 levels of automated driving. Unfortunately, this framework is not very useful because most of the interest lies in just 2 of the six levels, because it can be misinterpreted as conveying a linear progression of technology from level to level and because it is based on a limited, somewhat mechanistic perspective but fails to see the full complexity of the software-based self-driving vehicle and the complexity of the context in which it operates, which it interacts with and constantly learns about.

Fortunately, the group decided against addressing highly automated first and fully automated driving only beginning in 2016 (see annotated working group document). Both topics will now be considered somewhat in parallel, although the group still leans more toward highly automated driving. One of their future discussion items will be usage scenarios for highly automated driving. Maybe they will also consider some scenarios for fully automated driving and then begin to understand the extent to which mobility and with it the role of passenger vehicles will change.
An excellent source for information about this process is GlobalAutoReqs.com, which maintains an up-to date list of cross-referenced documents related WP9.

 

First fully autonomous Audi expected by 2017

Several news media have reported that Stefan Moser, Audi Head of Product and Technology Communications, has announced that the next generation Audi A8 (expected by 2017) will be able to drive with full autonomy. Mr. Moser emphasized that Audi wants to be first to bring a self-driving car to market. He explained that the car will be equipped with cameras and LIDAR, that the car will drive much safer than humans could, and that their system will be based on a redundant hardware architecture where all computing will be performed by at least two independent processors. He also cautioned that legal hurdles remain for fully autonomous driving which could delay the availability of these features.

This announcement shows that car makers increasingly want to be seen as innovation leaders in the autonomous driving space. Audi has a mixed record in this area. They have have been very active in the field of driving dynamics – i.e. racing a self-driving car up Pikes Peak or around the Hockenheim race track. But the sensing and route planning algorithms of these prototypes are still quite primitive – they rely mostly on differential GPS supplemented with custom-built 3D maps for navigation. Audi has made great progress in autonomous racing on empty tracks  but driving in a dynamic, changing environment with other vehicles, pedestrians, etc. is a different ball game. It does not help that Volkswagen’s CEO Martin Winterkorn remains quite sceptical about fully autonomous technology (Audi is a subsidiary of Volkswagen). On the other hand, Audi has established itself as a technology-leader with respect to the computing platform for driver-assistance systems via its partnership with NVIDIA.

We hope that Mr. Moser’s statements are an indication of a change of heart within Volkswagen and that they will aggressively tackle the challenges of autonomous urban and highway driving. This requires an extensive program of computer-based learning and optimization and needs millions of kilometers of test-driving with autonomous car prototypes on regular roads.

Source: Motoring.com.au, CarAdvice

Five guiding principles for autonomous vehicle policy

As self-driving car technology matures, politicians and regulators find themselves called to action. But the technology is a moving target and views about the technology’s path and impact vary widely. So how should policy makers approach the subject? Here are five guiding principles proposed by Marc Scribner,  a transportation and telecommunications policy specialist and research fellow at the Competitive Enterprise Institute. Scribner only discussed the principles briefly at a recent presentation at the Cato Institute. In the following I supplement each of his five bullet points with my interpretation:

1. Recognize and promote the huge potential benefits of self-driving cars

Policy makers need to familiarize themselves with the potential benefits of self-driving cars. First, they need to get the concepts right and clearly distinguish self-driving cars (which can drive without human supervision, even empty, and don’t need additional infrastructure) from other technologies such as driver assistance systems and connected cars. Connected cars and driver assistance systems are certainly also interesting topics but their benefits pale in comparison to the benefits of cars that drive themselves. Besides greatly reducing accidents, self-driving cars also bring individual motorized mobility to those who do not have a driver’s license – including people with disabilities and the elderly. They reduce energy consumption, simplify the introduction of alternative fuels and reduce the load on the road infrastructure.
Policy makers need to recognize that self-driving cars can solve or greatly reduce many longstanding problems. This is not a technology where a wait-and-see attitude is warranted. Politicians need to actively promote this technology. Of course, this does not mean that the technology’s risk should be ignored.

2. Reject the precautionary principle

Safety is a key concern and a key benefit of self-driving cars. There is good reason to expect mature self-driving cars to drive much safer than humans. They are equipped with 360 degree sensors, including cameras, radar and Lidar, are always alert, never tired, don’t drink and adopt a defensive, risk-minimizing driving strategy. But letting the first such cars drive by themselves on public streets is a difficult decision: what if anything goes wrong?
The application of the precautionary principle avoids this situation by requiring the developer to prove that the car is harmless. Unfortunately, proving that a self-driving car is safe is a hard problem and strict application of the principle could significantly delay the introduction of self-driving vehicles.
This weakness of the precautionary principle is well-known: There is the risk that erring on the side of caution when certifying self-driving cars prolongs the current carnage on our  on our roads. Unfortunately, we don’t have the luxury to delay a well-functioning self-driving car for a few more years to be extra-sure that everything is perfect when 33,000 people die in traffic accidents per year in the US alone and more than 1 million per year worldwide.
As much as it is not acceptable to let first prototypes roam the streets unsupervised it is not acceptable to delay and delay just to be on the safe side. A middle ground must be found. This is not an easy task for policy makers but one on which lives depend.

3. Don’t presume to know how the technology and law will evolve

Will autonomous vehicle technology gradually evolve from driver assistance systems? Will they first appear on the highway or in low-speed local settings? What new business models will emerge and what role will machines play? Will the US be the first to legalize fully autonomous vehicles or does the Vienna Convention on road traffic really prevent many European Countries from adopting self-driving vehicles? There are so many paths that this technology can take, so many changes in many different areas of business and society, so many proponents and possibly opponents that it is hard to be right about the path of technology and – consequently – of law. It is very dangerous to assume that the technology will evolve in one way, then regulate for this situation and subsequently find that the technology evolves very differently.

4. Let the innovators innovate

This section was originally entitled ‘minimize legislative and regulatory intervention’ and included the goal to give the innovators the space to innovate. But here I differ with Scribner: Unfortunately, transportation law is so much based on the concept of vehicles driven by humans that many laws do need to be changed. Current traffic laws contain so many elements that inhibit progress for this new and safer technology. Autonomous vehicles change the concept of what a car is and the laws need to be updated accordingly. Otherwise innovators will find it hard to make progress. This is a task that should be started immediately – before fully autonomous vehicles are ready for public roads.

5. Preserve technology neutrality

Laws and regulations should be technologically neutral. As much possible, they should avoid favoring a specific technical approach.

Autonomous vehicle roadmap: 2015-2030

Two and a half years ago I wrote a note on the various views about the paths for adopting self-driving vehicles. Since then, more and more signs point towards my ‘avalanche’ model, where the adoption of self-driving cars becomes a self-sustaining, accelerating process fueled by expectations of a fundamental transformation of the auto industry and major opportunities for profit.

As a thought exercise, I have sketched a hypothetical timeline which shows how this self-accelerating global innovation process could unfold. The purpose of the timeline is to show how autonomous vehicles could come into widespread use rather quickly and what kind of market and political forces could be involved. This is an extreme of many possible futures for self-driving cars:

2015 Google launches first short-range fully autonomous vehicle service in California at NASA Ames (not on public roads) and possibly in Mountain View (small scale pilot, limited to Google employees).

2015 The first auto makers (Daimler, Honda, Nissan?) announce major strategic initiatives and major investments to counter Googles’ threat and rapidly bring vehicles capable of full autonomy (Level 4) to the market.

2015 Car2Go (Daimler’s shared mobility service) announces a roadmap for autonomy in their car fleet.

2015 Automotive industry recognizes the implications of fully autonomous vehicles (transformation of mobility, significantly lowered worldwide demand). Analysts pound auto makers on their Level-4 autonomous vehicle strategy. Share prices begin a long decline.

2016 Google announces that their short range, limited-speed fully autonomous vehicle fleet will be built by Ford, Magna or others.

2016 China launches a major program to develop and deploy shared autonomous vehicles for local mobility. It recognizes that it can reduce infrastructure expenditure, jump-start their autonomous vehicle industry, reduce the ecological footprint of mobility etc.

2016 Google expands their short range autonomous vehicle service pilot to another US city that sees little rain and no snow, e.g. Las Vegas, NV or Sun City, AZ and starts their first overseas fleet.

2016 Price for semiconductor lasers used in LIDAR sensors falls below USD 150; this reduces the hardware/computing power costs for autonomous vehicles with 3D Lidars to below 10,000 USD.

2016 Transformative potential and benefits of autonomous vehicle technology are recognized widely. There is a bitter debate about the destruction of jobs.

2017 Several European countries have now adjusted their laws to allow the operation of fully autonomous vehicles on a national scale (not in international traffic).

2017 Autonomous long haul highway trucks start testing in the US, Europe or Japan.

2017 Rental car companies launch their own autonomous mobility inititiative.

2017 An international body for regulating autonomous vehicles is being formed in cooperation between the US, Europe and Japan.

2017 Google vehicles are now capable of driving in snow on pre-mapped routes.

2017 Automotive suppliers (Continental, Bosch, Valeo, or others) announce their own autonomous vehicles or special-purpose autonomous machines.

2017 Major road infrastructure projects are downsized because autonomous and connected vehicle technology have reduced the expectations on future transportation demands.

2017 Google moves their autonomous vehicle operations into a subsidiary which then merges with Uber and starts to roll out local autonomous vehicle mobility services in many more US cities.

2017 Singapore deploys the first autonomous bus for regular service. This is widely seen as a milestone for public transport and sends many transit corporations scrambling to update their strategies.

2017 The first countries mandate specific driving behavior for self-driving cars in certain driving situations.

2018 Car2Go starts to add autonomous vehicles to their fleet.

2018 The Google subsidiary/Uber merger rolls out autonomous vehicles internationally.

2018 Heavy investment into autonomous vehicle fleets and services based on autonomous vehicles. An almost unlimited amount of capital flows into startups and schemes. Countries compete trying to gain an advantage in the emerging new industries.

2018 Experience with autonomous vehicles shows that they are indeed much safer than the average human driver. People feel safe and comfortable in fully autonomous vehicles and there is no longer any question of user acceptance. No phenomenon similar to the ‘fear of flying’ can be found among users of self-driving cars.

2019 The Vienna Convention and European Laws are updated to allow the operation of fully autonomous vehicles.

2019 Autonomous vehicles now operate in over 50 cities worldwide.

2019 Rapid growth for autonomous trucks on specific routes. In many countries, truck drivers protest but this can only delay their adoption slightly.

2019 The first high-end consumer cars capable of fully autonomous driving on a large part of the national road network become available.

2020 The first countries introduce laws that prohibit bullying of autonomous vehicles (e.g. jumping in front of it to make it stop).

2020 Bleak outlook for automobile companies. Volume is down, consumers prepare for the transitioning to fully autonomous vehicles (which are not yet widely available for the consumer) or increasingly use/expect to use shared autonomous vehicle services. The fight for survival has begun: The auto industry has its “Kodak moment”.

2022 Prices for used cars decline. Too many people switch to shared autonomous vehicle schemes. Many others sell their old vehicles prematurely because they want to switch to the much safer fully autonomous models where they don’t need to drive if they don’t want to.

2022 The cost for autonomous vehicle hardware (sensors and computing power) has come down to 1500 USD.

2022 Mass transit companies increasingly rely on autonomous vehicles for transport. Transitioning the current workforce to a transit system based on autonomous vehicles is a major organizational and political challenge.

2022 Insurance rates favor operating cars in fully autonomous mode and prompt many people to stop driving on their own.

2023 Small autonomous buses are increasingly used for medium- and long distance trips. Trains have a hard time to compete on short to medium distances with autonomous buses.

2023 Most companies require that business trips with rental cars must occur in fully autonomous mode (for safety and productivity reasons).

2025 Fleets of autonomous vehicles now operate in most cities of developed nations.

2025 Automotive companies shut down more and more plants. Major automotive countries including Germany, Sweden and Japan desperately try to prop up their OEMs.

2030 Car ownership has declined dramatically. Only 20% of the US population still own a car (200 cars for 1000 people, today: 439 cars for 1000 people).  90% of all trips now happen in fully autonomous mode. Traffic accidents and fatalities have declined dramatically.