Autonomous vehicles¹ have made significant inroads over the last few years. You would not need a crystal ball to guess that there will soon be one in a garage near to you. Actually, they may not even need one! Critics and cynics may raise concerns, and indeed a few accidents have happened recently, but the statistics show that autonomous vehicles already drive much safer than humans. As the percentage of self-driving cars increases, the track record will improve. The US Department of Transportation aims at eliminating all fatal car accidents in the US 30 years from now.
This article gives an outline of the opportunities, risks and the reasons for why the process may take slightly longer time than some of the optimists claim. Fully autonomous vehicles will ultimately change our entire society, and we need to be ready for what may happen over the coming years and jointly shape our future.
Levels of self-driving
It may take ten years until fully autonomous cars are on our roads. On the other hand, such vehicles may then be the most common ones. However, cars with increasing degree of automation will become the standard much before that. Indeed, the development of self-driving cars with increasing levels of automation are already well underway. SAE International has defined six levels of autonomous driving.
- Level 0 – No automation
- Level 1 – Driver Assistance: Cruise control, lane guidance and automated parallel parking. The driver is still fully engaged in all operations.
- Level 2 – Partial Automation: Adaptive cruise control with lane centring. The driver is still responsible for monitoring and should be available to take over control at any time but may not need to touch the steering wheel or pedals.
- Level 3 – Conditional Automation: Drivers can cede all safety-critical functions under certain conditions. The driver is not expected to check the roadway. However, the car may ask the driver to take over under some circumstances. At this level, the driver can work while travelling on a motorway. The reaction time for a human to take over would be significant compared to the level 2 scenarios. This takeover time is a reason which speaks against this level of automation. Perhaps car manufacturers will have to skip this level and move directly to Level 4.
- Level 4 – High Automation: Fully Automated Driving System where all aspects of the driving are managed by the car, even if the human driver does not respond appropriately to a request to intervene. On this level, the driver can work in most conditions and perhaps even sleep when going on a motorway.
- Level 5 – Full Automation: The vehicle is fully automated and can manage all aspects of driving on any road and under any environmental conditions without any human intervention. Cars capable of Level 5 automation are likely to be around by 2030.
Autonomous cars are safe
85 thousand people die on the road every year in Europe. In the world, the figure is about 1.25 million. In other words, more than two people die on the road every minute. Forecasts say that within 30 years’ self-driven cars would cut that figure by 90%.
In May 2016, a new self-driving Tesla car collided with a tractor-trailer at an intersection on a highway in Florida, The driver of the Tesla died. However, the U.S. National Highway Traffic Safety Administration found no fault of the Tesla Autosteer technology. Instead, it found that Tesla cars were 40% less likely to crash when driving in self-driving mode than when not. However, Tesla would immediately improve its autopilot system by sending a software update to all cars. The next time a similar situation happens, any Tesla car will be able to manage the situation and avoid the accident. Unlike what is the case for human drivers, self-driven cars can learn all the time, and share the knowledge across the entire fleet of Tesla cars. By 2050 as many as 30 thousand lives can be saved every year, thanks to self-driving cars.
In Sweden, there are between 45 and 60 thousand reported accidents involving wild animals every year. Human drivers find it hard to see approaching animals, especially in the twilight. In the majority of these accidents, only animals are injured, and the vehicle may be damaged. Human injuries are as low as about 800 per year, and there are less than ten fatal accidents per year. Volvo Cars in Sweden is developing technology which will find out when wild animals are about to cross the road and can avoid accidents. Volvo claims that by 2020 nobody will be killed or seriously injured in a new Volvo.
We can assume that an autonomous vehicle can assess a situation in about 0.3 seconds. Humans, however, may have a reaction time of two seconds. Assume that two cars both drive 90 km/h of which one is autonomous, and the other has a human driver. Given the reaction times above, the autonomous vehicle would apply breaks or divert after only 7.5 meters against 50 meters for the human driver. The autonomous vehicle would also better take into consideration other vehicles and be far better at avoiding cars behind driven by human drivers. In this way, autonomous cars can obviously save lives. Note that this is possible even with Level 2-3 automation.
Autonomous cars such as Tesla, BMW, Mercedes and Volvo will have their early adopters, especially in the luxury segment. However, the real commercial gain may not be in self-driven private cars but autonomous trucks and buses. Here are some reasons why we can expect transportation companies to adopt autonomous vehicles:
- Salary costs
- Ability to run 24×7
- Labour legislation
- Fewer accidents
- Less risk for theft of goods
Long haul drivers tend to work long working weeks and fatigue is one likely reason for accidents. The technology is already around. Initially, transport companies may aim at reducing the number of drivers to one per vehicle and later entirely remove the drivers. To give an idea of the scale of the impact of this – In Europe alone there are about 13 million heavy-duty vehicles and 23 million light-commercial vehicles.
New visions for transportation
There are different views on how future transport will develop. Will transportation be done by small vehicles directly from production to destination or will large trucks still be used for long-distance transport?
Likely, there will not be one solution and combined with new production methods (Automation, 3D printing, etc.) we can expect other changes to the requirements for future transport as well. Since most manufacturing may move closer to the where the consumers live, smaller autonomous vehicles (or even drones!) may make more sense than larger vehicles. On the other hand, where long haul transport is required, larger automated trucks between hubs may make more sense from an economic, environmental perspective.
There is a lot which speaks for this transformation;
- Reduced number of accidents
- Totally remove the risk of fatal accidents
- Ability to work in the car
- More people can share a car since the car does not need a driver to move between places
- Car sharing pools
- The company provided shared vehicles on demand (Such as Uber)
- Reduced cost of transport due to no need for drivers
- Environmental gains due to more optimal driving techniques
This transformation will, however, be major social impacts on society. Some of these changes may be good, but others may if not mitigated create hugely negative societal impact. It is likely that many drivers will lose their jobs. The most common profession in many US states is the truck driver, and millions of people will lose their job over time. This change will take years, and not all trucks will be replaced at once. Some of the current drivers may retire before they are made redundant by an autonomous vehicle. But it makes sense to assume that within 20 years all long haul trucks would be fully autonomous.
What can delay autonomous cars?
The opportunities are enormous so when will this revolution take place? There are some factors which can delay or affect the roll-out of autonomous vehicles. The main ones are probably regulation and compliance, improving the Human-Machine Interaction, getting people to trust the cars and to make sure that cars can manage extreme road and weather conditions.
Policy and regulation
So far only a few US states allow testing of fully autonomous vehicles on roads. However, far more cars are driving in a semi-automatic mode where the driver still has his or her hands on the wheel. The driver can at any time take over and may legally be responsible for the actions of the car. The Tesla accident mentioned above was such a case.
Policymakers may be hesitant to implement new legislation, and such delays may slow down rolling out of autonomous cars. At the same time, given the dramatic impact on road safety, it is safe to assume that driver assistance will become mandatory. In the US all cars must have automatic breaking facilities from 2022.
Insurance and liability
Even though accidents will be less common with autonomous vehicles, insurance companies may argue whose fault an accident is. However, car manufacturers will take full responsibility as long as the car is in autonomous mode. Hence this is not likely to become any major obstacle. German authorities have already started the work to define ethical rules for autonomous cars. They are defining how the prioritization between human lives, animals, and other things should be done in a potential accident scenario. This is not something human drivers normally think about. This may be a policy issue but is closely connected with liability.
Since autonomous cars use navigation equipment and communicate with servers and possibly even with other vehicles, data about where each person car is located would be collected in real-time. It would be possible for car manufacturers to track and predict standard routes for each person. Government agencies and companies may use such information to track cars or for monitoring, advertisements and marketing.
However, since most of us already use smartphones, this information is already known to them. Telecom operators can also, obviously not with the same precision, track in which area we are. Tracking cars may not make a significant difference. Government and Corporations can always potentially use and misuse of collected information. It is possible to anonymise data as well as implement regulations and compulsory privacy audits. But tracking autonomous vehicle is only one more risk.
If autonomous cars are connected to the Internet, there are risks that they can be hacked. Hacking may be done to find out where certain people or cars are, but rogue hackers could also take over one or many cars. Potentially high-tech terrorists can threaten civil society in novel ways. Vehicle manufacturers must make sure that communication is secure and that various protective systems are implemented to track unexpected deviations and to react to suspicious situations.
Such solutions are already available in general IT systems. However, unless the software of all included devices is kept up to date, there is always a risk. As the vehicle park ages, manufacturers may stop supporting old versions. This lifetime of an autonomous vehicle may, therefore, be shorter than a traditional car. This is similar to when Microsoft stops supporting old operating systems.
How will society change?
IHS forecasts that 76 million autonomous vehicles will be sold over the next 20 years, which is about the same number of cars that were sold worldwide during 2016. That may not sound impressive, but since one single autonomous vehicle can potentially replace nine owner-driven vehicles, the figures may imply a far more radical transformation than the statistics imply. How will our society change as a result?
Our arterial roads are getting more and more congested. Instead of constantly adding more flyovers and widening the roads, autonomous cars will be able to drive closer to each other and we will be able to get more cars into the same space than with human drivers.
Cars with automation level 1-4 are for all practical purposes nothing but an incremental improvement of existing owner-driven cars. However, level 3-4 would still mean that the driver can work during some or all parts of the trip, but there is still a need for a person with a driving license. Only cars with level 5 imply a fully self-driven car, All levels of automation will indeed make driving safer, but only level 5 will be truly transformational. For level 5 cars there may not be any need for a steering wheel or pedals or a person with a driving license.
The only thing we can be sure of is that all our prediction about the future will be wrong (including this one). The emergence of self-driving cars as such is already well on the way, and we can reasonably well predict the development by just extrapolation of today’s development. However, while the technical capabilities can easily be predicted the changes in social patterns and business models are harder to predict.
Ten years from now, most new cars are likely to be fully autonomous driving. However, since the average lifespan of a car is 20-25 years, it will take almost 20 years from now until the full effect of this change would be experienced. Since many people may lose their jobs as a result of increased automation, governments must consider what to do. Stopping the technical development is likely not to be a relevant way. Elon Musk founder of Tesla, SpaceX and Hyperloop suggests universal basic income as one possible solution.
New patterns of transportation
Since autonomous cars can run 24×7 car, it would be a wasted opportunity to keep an autonomous car be unused when it could be used. Therefore, it can be predicted that other business models than traditional ownership will emerge. Self-driving cars may be tomorrows near distance public transportation.
Obviously this is going to change in steps, and initially self-driven cars would probably be owned by people, exactly as today’s cars. However, over time, when self-driven cars can be used around the clock, the number of cars required would go down. Welcome to Car-as-a-Service – CaaS! Manufacturers, Uber and Lyft already plans for such solutions!
Think of a perfectly AI-driven optimised taxi system without drivers. A system of this sort will compete with some traditional public transportation. In particular, slow, inefficient types such as buses and trams. Existing underground systems, especially in larger cities, may be able to retain their position initially.
Over time, fully autonomous vehicles will be replacing traditional owner-driven cars and congestion is likely to be reduced. In the long run, autonomous vehicles working together with coordinated AI systems will reduce delays and congestion. Thanks to this we may change how we build cities and unless now the node-based public transportation system may not any longer be the primary factor for where buildings and offices are built. Such societal changes may even, in the long run, reduce the need for the underground. Ultimately fully automated cars may be the only required public transportation system.
Longer range travel
For long-range travel, some node-based extremely fast new technologies such as HyperLoop may take over from trains and flights, especially on high volume corridors between larger cities. However, Airbus and Siemens are planning to develop a small electric aircraft with less than 20 seats for short-range transportation.
It does not take much imagination to predict that these may become autonomous and may not even need an airport. Environmental objectives will be a major driver, but since the gains would not be limited to the environment but could generate significant benefits in the form of less congestion, shorter and more convenient travel time, it is likely that this will sooner or later be driven by other objectives as well. The real transformation will take place when autonomous vehicles reach automation level 5. However, then we can expect massive changes to our entire society. Before that, automation will just be a way of supporting humans and make road travel safer and more environment-friendly.
Moving away from the need for a hub-based system of local transportation will have far-reaching consequences for how our societies will be organised. A city centre will not necessarily be required. Will we perhaps prefer to live in villages again? Nobody knows. We can expect a massive revolution, and many questions remain unanswered today.
The next few years will be fascinating. Many manufacturers are about to introduce new self-driven cars; legislation is likely to be implemented in more countries. We are likely to see an explosion in the coming years. There are still a few obstacles on the horizon. Some of them are relatively easy resolved while others may be more challenging. In particular, the social issues of people losing their jobs are concerning.
Some countries may hesitate to implement the required new legislation. Reduced number of accidents, commercial gains and increased comfort will ultimately make it happen. New business models where cars would no longer be a product but a service are likely to emerge. How will this change our society? Where and how we will live, work and interact?
We are at the edge of a revolution on par with the industrial revolution, and nobody knows what will happen. It is evident that road safety will improve even with lower levels of automation, but what will happen next remains unclear. Maybe the question when autonomous cars will take over is the wrong question. Perhaps we instead should ask ourselves when humans would no longer be allowed to drive cars at all?
Mikael Gislén is the Managing Director of Gislen Software, a Swedish-owned Indian Software development company. Mikael started the company in 1994. Gislen Software provides high-value software development services to clients mainly in Scandinavia and the UK.