Autonomous cars : will roads be safer if algorithms replace human drivers?

A boost for research from the European Parliament

A public consultation on robotics and Artificial Intelligence consultation coordinated in 2017 by the European Parliament covered, among other things, also autonomous vehicles. Based on the results of this consultation autonomous vehicles have been identified as a policy area requiring most urgent action.

Public consultation on robotics and Artificial Intelligence by Dr. Tatjana Evas European Added Value Unit 13 July 2017

The first EP legislative resolution on the topic (2017) ‘Civil Law Rules on Robotics’ emphasizes that the switch to autonomous vehicles will have an impact on the following aspects: civil responsibility (liability and insurance), road safety, all topics related to environment (e.g. energy efficiency, use of renewable technologies and energy sources), issues related to data (e.g. access to data, protection of data, privacy and sharing of data), issues related to ICT infrastructure (e.g. high density of efficient and reliable communication) and employment (e.g. creation and losses of jobs, training of heavy goods vehicles drivers for the use of automated vehicles); emphasises that substantial investments in roads, energy and ICT infrastructure will be required

On January 15th 2019, the European Parliament has approved a motion calling for greater development of and more funding for research on automated vehicles, while members also emphasised the need to establish appropriate rules regarding safety and liability. The stakes are high, because, as rapporteur Wim Van de Camp (EPP; NL) said: “Europe has to be innovative, but faster. China and the USA are not waiting.”

We are at the testing phase

Companies from the United States’ Silicon Valley and tech businesses funded by the Chinese government have been testing autonomous cars for several years now, but the public opinion surrounding them is significantly different in the two countries: according to a survey released in February 2018, only 34% of US drivers believe driverless cars will increase road safety, compared to 63% of Chinese drivers.
The basic idea behind autonomous cars is that, by removing the human factor and human error from driving, they will increase safety, mobility and cost-efficiency. At the same time, there are numerous ethical, technological and liability issues to consider. As it is an emerging technology, for which it is very difficult to obtain permissions to test in real-life conditions, policymakers have to make decisions based on projections and assumptions, though lately it has been changing a lot. There are more and more countries and cities that allow public pilot testing of autonomous vehicles.

Public online survey on automated driving

An EU funded research project „Autopilot” brings together relevant knowledge and technology from the automotive and the Internet of Things (IoT) value chains in order to develop IoT-architectures and platforms which will bring automated driving towards a new dimension. An international public online survey has just been launched.
One of the most intriguing questions decision-makers face is that we do not know whether autonomous vehicles will eliminate all human-driven cars, or they will coexist, and we are also not certain whether people will own their private self-driving cars, or transportation will be entirely service based. Alternatively, these two business models may be available at the same time.

Jack Stilgoe, associate professor at University College London, fellow of the Alan Turing Institute: “At the moment, all projections are utopian. The technology, in its nascent form, has only just been introduced to the real world and has barely got to know all of the constraints and foibles of real life. Car companies will want to keep selling cars to customers, and some people will fight hard for their right to drive. The lobbyists for new technology may not be able to push them all out of the way.”

But is it possible to make decisions without the technology in discussion even entirely existing? Fortunately, there are many fields of technology needed for a transport system based on autonomous vehicles that have seen significant development over the last few years. Beside artificial intelligence that feed on incredibly large datasets assembled by humans, deep learning algorithms are also rapidly becoming more powerful. AlphaZero, for example, which learned to play chess from scratch to world champion level in nine hours, was a product of Google’s parent company, Alphabet, that also owns one of the world’s largest autonomous car developers, Waymo. Today’s autonomous cars use deep neural networks to learn from datasets based on real-life driving scenarios, which help them calculate the best course of action in any given traffic situation.

Advanced mechatronics have also helped pave the way for autonomous cars, so that many different technologies can be combined into high-tech systems which form the basis of self-driving vehicles. These technologies include detection, surveying and navigation systems such as radar, lidar, sonar, digital cameras and GPS. With deep learning algorithms, these technologies are responsible for everything a human driver needs to pay attention to, including identifying objects and other traffic participants.

The six levels of automation established by the Society of Automotive Engineers (SAE)
• Level 0 : Fully manual vehicle
• Level 1 : The automated system and the driver share control of the vehicle (Adaptive Cruise Control, Parking Assistance or Lane Keeping Assistance)
• Level 2 : The automated system takes full control of the vehicle, but the driver must be prepared to intervene at any time
• Level 3 : The automated system handles situations that call for an immediate response as well, but the driver must be prepared to intervene within limited time (driver can do other activities)
• Level 4 : No driver attention is needed for safety, except outside of geofenced areas (driver can sleep and leave the driver’s seat)
• Level 5 : No human intervention is required at all

However, the aforementioned technologies also raise a significant ethical dilemma: when autonomous cars will be able to tell the difference between men, women, adults, children or animals, whose life should they decide to spare in the event of an unavoidable crash? This contemporary equivalent of the famous trolley problem was examined in the grandiose online quiz by MIT – the Moral Machine experiment gathered more than 40 million decisions from millions of people in 233 countries. Researchers tested nine separate factors, based on the age, gender, social status or number of the people involved in the imaginary accidents.
The authors of the study based on the MIT experiment found some consistent global preferences: people would spare humans over animals, children over adults and more lives rather than fewer. The top four and the bottom three of the global priority list were selected by respondents by a wide margin: people would not hesitate to save babies in strollers, girls, boys and pregnant women, while they would rather crash into dogs, criminals and cats than others.

Patrick Lin, director of the Ethics + Emerging Sciences Group at California Polytechnic State University: “We already prioritize life in hospital triages, organ transplants, and so on. But those are drastic cases where there’s no other option. We’re not even close to being there yet in the conversation about autonomous cars. Giving people a priority score is not like giving them a credit score – it directly affects their right to dignity as well as right to life. Not every state cares about these things, of course, but luckily they tend to not be the ones developing the technology either.”

To deal with these issues, in 2018, the European Group on Ethics in Science and New Technologies proposed a set of basic principles in accordance with fundamental European values, such as human dignity, rule of law and data protection, but one of them tends to come up the most often when it comes to the “moral principles” followed by autonomous vehicles: the issue of responsibility. Who should design the decision-making process of self-driving cars: programmers, companies or governments? According to Patrick Lin, it seems like every stakeholder will have a say in it: “As it is right now, product development is generally left to the company, which responds to market pressures. Sometimes in the case of drugs, vehicles, and other products that could be dangerous, regulators become involved, and they respond to public pressures. This arrangement seems correct.”

Although the policy-making process has already begun, we could still be far from the time when cars driven by algorithms take over the streets.

Jack Stilgoe concludes that : “the uptake of self-driving cars at scale will require substantial changes to behaviours, infrastructures and the rules of the road, as well as improvements in the technology inside the cars, which means that the transition will be much slower than the enthusiasts predict.”

However far it may happen, several companies are already deep into the development of autonomous cars. Big players like Google, Uber, Tesla or Baidu lead the way, but other tech giants, such as Apple, Huawei or Nvidia also got in the business, while large car companies have paired up with tech firms or part suppliers (BMW with Intel, Mercedes with Bosch, Hyundai with Cisco), and even competing manufacturers like Ford and Volkswagen form alliances to become leading forces in the autonomous car market.

Useful links
Call for transport projects coordinate by European Commission – Horizon 2020
Self-driving cars in the EU: from science fiction to reality
EPRS study on “A common EU approach to liability rules and insurance for connected and autonomous vehicles”
New 5G cross-border corridors for connected and automated mobility in the Baltics will allow testing of autonomous vehicles