Traffic fatalities are the leading cause of premature death of people under 30 in the US. Most are pedestrians and cyclists. The question is how to redesign traffic to reduce the number of victims.
Any human activity that causes 1.35 million deaths a year globally, more than 20 million injuries and a total damage of $1,600 billion and is a major cause of global warming would be prohibited at once. With the exception of traffic, as it is tightly connected with our way of life and commercial interests.
Most citizens are used to the ubiquity of cars, cherish their benefits and become immune to their costs, which are about 20% of average family income. The infrastructure of cities worldwide is dominated by cars, which is not surprising given that investments in roads for cars world-wide are $1,000 billion per year. Without intervention, global car ownership and use will growth exponentially over the next 30 years. Still, cars are stationary 96% of the time.
For this reason, ‘Vision Zero cities’ such as Oslo and Helsinki are committed to reducing the number of fatalities in traffic to zero in the next decade.
Vision zero cities is part fifteen of a series of essays on how cities can become more humane. That means finding a balance between sustainability, social justice and quality of life. This requires far-reaching choices. Once these choices have been made, it goes without saying that we use smart technologies to achieve these goals.
The essays that have already been published can be found here.
Human suffering caused by road incidents is all the more unacceptable because the causes are known: Driving fast and road design. The use of alcohol and drugs, fatigue and illness are mediating variables. Since changing driver behavior so far proved to be difficult, the main question is whether changes the design of roads and cars themselves can provide a solution.
To answer this question, this article consists of two parts. The first part explores how improvements in road design can enforce behavioral change of road users and mitigate the effects of accidents. The second part is focusing on cars and explores whether ‘self-driving’ cars will contribute to safety.
Road design and safety
Researchers from universities in the US, Australia and Europe have studied the relationship between the road pattern, other characteristics of the traffic and the road safety. They compared road design of nearly 1,700 cities around the world with data on accidents, injuries and road deaths. The study divided these cities into nine groups (see map below). Lead researcher Jason Thompson concluded: It’s quite clear that places that tended to have more available public transit, particularly rail transit, tended to have lower incidents of injuries.
The results were published in Lancet Planetary Health. Decreasing car use and increasing use of public transport, walking, and cycling are the main drivers of improving the liveability of cities in general too.
As shown in the map below, these characteristics apply primarily to European cities. Major US cities have also expanded public transportation, but the streets are dominated by cars, leaving less space for other road users.
Protection of weakest road users
Most accidents occur in developing and emerging countries. Road deaths in developed countries are decreasing. In the US from 55,000 in 1970 to 40,000 in 2017. The main reason is that cars better protect their passengers. This does apply to collisions with pedestrians and bicyclists, including many children. Their number is increasing significantly; in the US more than in any other developed country. The number of bicycle lanes has increased, but changes to the layout of the rest of roads and the speed of the motorized traffic have been left behind, exposing cyclists to the proximity of fast-moving or parking cars. SUVs in particular turn out to be killers and their numbers are growing fast.
Vision Zero Street Design Standard
No fatalities occurred in Helsinki and Oslo in 2019. These and other cities use the Vision Zero Street Design Standard, a guide to planning, designing and building streets that save lives. Accidents are often the result of fast driving, but are actually caused by roads that allow and provoke fast driving. To qualify as a Vision Zero design, three conditions must be met:
- Discourage speed through design.
- Stimulate walking, cycling and public transportation.
- Ensure accessibility for everyone, regardless of age and physical ability.
These conditions result in ten principles of Vision Zero, which bring safety in the fabric of streets and thus condition drivers for safe behavior and prevent errors from leading to accidents (see illustration below).
Separation of road users
Road construction contributes to the prevention of accidents. Major improvements have been made in recent decades, particularly in European cities. One of the most effective measures is the construction of separate lanes for all categories of traffic. Sidewalks have been around for years, but in many cases have become too narrow, due to the increased number of users. Over the past few decades, free-standing bicycle and bus lanes have been constructed and all have contributed to safer traffic. However, all these free lanes have many intersections, and despite protection with traffic lights, they pose a potential risk, primarily because the overview is limited. In addition, cars often have to cross cycle paths to park or to deliver goods.
Other examples of interventions to make roads safer are roundabouts, mainly because they reduce speed and create more clarity. The same goes for speed bumps, which are in fact a punishment for people who drive too fast.
Bicycle lane construction
In many American cities, paint is the main material for bicycle path construction. Due to the proximity of the traffic, this type of cycle path contributes to the increasing number of fatalities. The Canadian city of Vancouver, which saw the number of bicycle trips double to 11.9 % in five years, has the ambition to upgrade the bicycle infrastructure 100% to a AAA-level (meaning ‘safe and comfortable for All Ages and Abilities). Cycling paths must be technical safe: at least 3 meters wide for bidirectional use; separated from other traffic, which would otherwise have to reduce speed to less than 30 km/h. Users should also feel save too. The city published all its experience in an online set of AAA design guidelines.
Speed and legislation
Driving speed risks are contextual and depend on the time it takes to stop a vehicle if conditions require. Therefore, in some countries driving at walking speed is mandatory in residential areas. Strict speed limits require compliance and law-enforcement; both are challenging. In the Netherlands drivers of passenger cars are fined on average once every 20,000 kilometers. In addition, there are many digital devices that warn of impending speed traps. Given the risks of driving too fast, and the frequency with which it happens, this inadequate enforcement is astonishing.
Technology and safety
The amount of electronics in cars has increased drastically in recent years. First catalysts, then ATB and other systems that keep cars on track, navigation and automatic parking. Developments that have laid the foundation for self-driving cars. The progress of car automation and its impact on safety are examined below.
The contribution of car automation to safety
When are cars theoretically completely safe, apart from being hacked?
- Their communication works flawlessly with their surroundings (buildings, other cars, bicycles, pedestrians, children, animals and objects), under all circumstances (day, night, city center, countryside, rain, fog or snow) thanks to their sensors, cameras, lidar, radar and GPS.
- Their mechanical systems can compensate for any form of defect.
- They use 3D maps, which are accurate to the centimeter and can be adapted in real time to changing conditions
- They learn from ‘mistakes’ of all other cars thanks to artificial intelligence.
- They adhere to the traffic rules and that never go faster than the prescribed speed.
- They have no human driver…..
There is a big difference in the degree of automation of ‘self-driving’ cars. It is enlightening to take note of the six-level classification developed by the Society of Automotive Engineers (SAE) (figure below).
According to this classification, SAE level 2 cars can be trusted to steer and accelerate automatically in specific circumstances such as motorways. Under these conditions, drivers can keep their hands off the wheel, provided authorization by national laws. As soon as the environment makes steering and accelerating more complex, for instance after entering a busy street, the driver has to take over immediately.
A well-functioning SAE level 3 system enables drivers to keep their eyes off the road and engaged in other activities. Again, the only condition is that they immediately take over driving as soon as ‘the system’ gives a disengagement signal, which means that it can no longer handle the situation.
For driverless taxi services, this level of mastery is not satisfactory. Automotive and technology companies like Lyft, Uber and Google are busy to qualify for higher levels. Their expensive cars (up to $ 250,000) have automated fallback mechanisms, meaning that they can handle any situation under specified conditions, such as well-designed roads, during the day and at a certain speed.
In view of this classification, it is preferable to use the term ‘self-driving car’ no longer, but to distinguish automated cars (SAE level 1,2 and 3) and autonomous cars (SAE level 4 and 5). In the first group facilities for manual driving are always available, and the driver must sit behind the wheel and be standby. The second group can drive without a driver – type 4 under certain conditions – but in most current prototypes steering devices are available for interventions by an accompanying ‘safety driver’.
While SAE level 5 is the ultimate ambition, the automotive industry is not hurrying to remove the driver from behind the wheel. Her mission for the next decade is to sell as many electric cars to private individuals as possible, with automatic systems as nice-to-have features. Therefore, consolidating SAE level 2 – 3 is first priority. The inbuilt functions will contribute to the safe use of cars if drivers use them cautiously. Recent research by Connecting Mobility reveals that car drivers use automated systems such as Lane Departure Warning, Emergency Brake or Adaptive Cruise Control that enable early recognition and avoidance of dangerous situations only every now and then.
At the other hand, technological companies like Google, Lyft and Uber cannot wait to reach SAE level 4, which opens the way to driverless taxi services, albeit under restrictive conditions (quite roads, daylight).
There are also differences in the way car producers improve the safety of the cars. One option is improving the quality of the perception of the environment. A step in this direction is integrating all qualities of the camera into the lidar. The other approach is continuous training of neural networks. This is the strategy Tesla follows instead of using a lidar. Al Tesla cars over the word add images that they cannot decipher and wrong predictions by the autopilot that are corrected by the driver to Tesla’s training set. In the short video below Andrej Karpathy, Tesla’s director of artificial intelligence, explains how this process works.
The last two years, speculation about an invasion of autonomous vehicles on the road has died down or as leading marketing company Guidehouse Research Leatherboard concludes: 2019 was a challenging year for the automated driving (AD) sector as the reality of the challenges it faces continued to sink in.
Automotive company leaders have become increasingly pessimistic about the pace of the implementation of automation and they are also pessimistic about the growth of the market for electric vehicles (EVs). Uber predicted to have 75.000 autonomous vehicles on the road in 2019, without the presence of a safety driver. There were zero. Instead, new CEO Dara Khosrowshahi announced that it will take more than 50 years for all Uber cars will be driverless. His colleague, ceo John Krafcik of Waymo expected that it would take decades for autonomous cars become widespread and that level 5 would likely be out of reach. Volkswagen’s ceo said that fully self-driving cars might ‘never happen’. This is in contrast to Elon Musk’s claims that all Tesla cars will reach level 5 in 2020, thanks to its new Full Self Driving Chip, which can perform 36 trillion operations per second. He seems to ignore the problem of the quality of the data input, given extremely changing environmental conditions.
The human brain far outperforms modern machine in its ability to evaluate complex navigational hazards on the road. Artificial intelligence is way faster, but its accuracy and adaptability need serious improvements before it can be deployed on its own. Driverless cars struggle with ‘randomness’, caused by children, pedestrians, cyclist and human drivers and also with potholes, detours, and worn markings, snow, rain, mist, and darkness.
Traffic lights too might be a problem; the whole world has seen an Uber car driving through the red light.
SAE Level 3. The real problem?
Until now, it seemed that the problem is to attain SAE safety level 5, but there is a growing realization that the SAE levels 2 – 3 are the real problem. It is suggested that drivers at SAE level 3 can read a book but must stay behind the wheel to take over in event of a ‘disengagement signal’. But drivers either pay attention, or they don’t: According to studies in controlled test environments the response time of drivers who do not pay attention before they become aware of the signal is too slow to avoid safety issues. Rather than the five-level classification of the SAE, it makes sense to distinguish between two options: Assisted driving or autonomous driving. Assisted driving (levels 2 – 3) basically requires the driver to be alert and ready to intervene. No book reading! The use of the term ‘auto pilot’ must be banned. U.S. Senator Edward Markey recently publicly rebuked Tesla and demanded that it “rebrand and remarket” this feature because of its potentially misleading name.
The automotive industry has so far invested € 250 billion in EV’s and AV’s, something that one of the ceos called ‘downpaying on nothing’. Here you will find an overview of the progress of the main ‘players’ in the automotive industry that intent bringing ‘self-driving’ cars to the market.
The reasons for the U-turn in the sentiment is obvious. As recently revealed, carmakers themselves never believed their own expectations but for marketing reasons they echoed each other. All autonomous car manufacturers realize that the required radar, lidar, AI computing, 3D maps (see here for an explanation) will make these cars unaffordable, unless they use their own lanes, have a limited speed of operate in quiet neighborhoods.
Due of these complications both Waymo and Cruise (GM) have delayed the start of their commercial taxi services.
Nevertheless, in a couple of places such as Detroit, Stockholm, Tallinn, Berlin and in the near future Amsterdam, driverless minibuses are on the road. They are small vehicles with a maximum speed of 25 km/h, driving in a free lane or in traffic-calmed streets and on a fixed route, accompanied by a test-driver for the time being.
Lessons from accidents
There have been a few fatalities with ‘self-driving cars’ that provide insight in their current weaknesses. On March 18, 2018, an Uber self-driving car hit a woman crossing the street with her bicycle. At that time, the mandatory ‘safety driver’ behind the wheel was looking at the diagnostic instruments, which was allowed. Less than a second before the crash the driver looked at the road and managed to slow down. Analyses afterwards revealed that the system identified the woman as ‘an unknown object’ six seconds before the crash, and attempted an emergency brake 4.7 seconds later, but the Uber team had disabled this feature.
Less than a week afterwards, a Tesla Model X changed direction without any need, hit a concrete barrier, caught fire, which killed the driver. As with another fatal accident involving a Tesla Model S, the driver had switched on the autopilot-system driving on a road where its use was prohibited and had also taken his hands from the wheel.
Automotive and technological companies have extensively tested autonomous cars. Waymo alone counts 9 million kilometers without severe accidents and in no accident the driverless car was in fault. In each of these kilometers, however, a safety driver was on board, who could catch any disengagement signal and prevent other harm.
If the accidents involving Uber and Tesla reveal anything about safety, it is that Uber’s Volvo cars are certainly not ready for autonomous driving at SAE level 4, apart from the irresponsible behavior of the team. Considering Tesla, excessive reliance on the TESLA autopilot (somewhere between SAE level 2 – 3) is deadly dangerous.
From the perspective of road regulation, automated or driver assisted vehicles (SAE Level 2 – 3, with a human driver) and autonomous or driverless vehicles (SAE Level 4 – 5, without a driver) differ fundamentally. Regulations regarding autonomous vehicles have so far prevented driver-less cars from appearing on the road.
The State of California recently proposed new rules to allow autonomous cars; several other states in the US followed. As a result, Google and General Motors have been allowed to launch driverless taxis services, albeit with the mandatory presence of a ‘safety driver’ during the pilot period. The trips will be monitored with cameras to prevent reckless behavior or vandalism. Testing these services is still on-going.
Until now, and in the (near?) future the development of autonomous cars is based on safe participation in traffic that is dominated by ‘ordinary’ cars and other users of the road. Real progress in safety will be achieved once driverless cars can communicate with each other and the presence of human-steered cars on public roads is prohibited, not to mention horse-drawn vehicles. This will last for decades, and in the meantime the automotive industry will focus on selling electric vehicles at SAE levels 2 and 3 for individual use.
Safety and the humane city
The large number of fatalities and injuries that are the price of traffic worldwide testifies to a serious lack of humanity. Given the many other aspect in which traffic impairs the quality of life in cities, changing mobility patterns have the highest priority.
Safety in traffic, as prescribed in the Vision Zero Street Design Standard makes a major contribution to the growth of humane cities. The leading causes of fatalities and severe injuries are driving too fast and the road design, combined with the use of drugs and alcohol, fatigue and illness. It is risky to expect that driver behavior change can solve this problem. Therefore, the underlying causes have to be addressed. This includes a fundamental redesign of the roads and adapting the speed limit to the main users of a road and enforcing this speed. The other measure is to improve car safety. Passenger safety has already improved significantly, but their design and speed have made cars an increasing threat to pedestrians and cyclists. Autonomous cars can theoretically be a benefit for safety, but their development stagnates.
In most cities, especially in the US, cars (or their drivers) have appropriated a disproportionable share of the available space. Pedestrians, cyclists, scooters, trams and busses struggle for the remainder. Therefore, authorities need to redistribute space in a way that reflects the changing mobility preferences of citizens and is aligned with sustainability, liveability and safety objectives.
Below I summarize the essence of a humane approach to safety in traffic.
Principles for a humane approach to safety in traffic
- Reducing the 1.35 million deaths and 20 million seriously injured every year as a global and underrated human disaster.
- Improving road design, ensuring safety for pedestrians and cyclists when designing the cars, and effective law enforcement should be a priority.
- The construction of different types of roads, each with its own speed limit, will make an important contribution to increasing road safety.
- Systems reporting speed traps, including P2P, should be prohibited.
- All cars must have a black box, which, among other things, records the speed. In the event of accidents or crimes, it can be consulted after approval by a judge.
- Reducing the number of cars, promoting walking or cycling, using public transport and decreasing the number and length of journeys will reduce the number of road accidents.
- In the city center, the space is mainly intended for pedestrians and cyclists. Cars are allowed to enter this area when driving at a walking pace to maintain accessibility to hotels.
- Separate cycle paths connect the main destinations within the city. For safety reasons, a maximum speed applies, unless there is room for different strips for slow and fast cyclists.
- Rather than the five-level SAE classification, it makes more sense to distinguish between two options – assisted or autonomous. In the case of ‘assisted driving’, the driver must always be attentive and be able to intervene immediately. The use of the term “autopilot” should be prohibited.
- In the long run, a system dominated by autonomous cars is not compatible with driven vehicles. At that time, these should be banned on roads intended for cars.
Written by Professor Herman van den Bosch, Professor at Open University of The Netherlands.
Header image: Motorbike security – SVG Public domain – Pixabay.