Autonomous cars. Test. How do they work? | Tomasz Leżoń, TVN24 Magazine

share

I'm approaching a corner, but I don't grab the steering wheel or apply the brakes. I am doing nothing. I think I'm about to get off the road and crash my car. The car, however, smoothly maneuvers and stops at a red light. These are just some of the benefits of autonomous vehicles, which are to become the world's main means of transport in the future. This is a real revolution that is coming faster than we think.

Ever since Henry Ford invented the moving assembly belt concept, the automotive industry has witnessed continual advances in technology that ultimately led to cleaner, inexpensive and safer vehicles.

But at the end of the second decade of the 21st century, the industry is facing a landmark change, perhaps the greatest in its history. Cars and the technology they use are set to change more in the next ten years than in the last three decades.

Revolution

For several years, the largest automotive concerns and technology companies have been investing billions of dollars in the development of autonomous cars, i.e. cars that can move without a driver. You might think that the world has gone crazy, but the revolution is happening here and now, before our eyes, although we are not fully aware of it.

Dreams of cars that would move without a driver have not been born recently. Already in the 1920s, a certain Francis Houdina presented a car controlled by radio on the streets of Manhattan. According to the New York Times report, the inventor started the engine in this way, changed gears and honked his horn, "as if there was a phantom hand behind the wheel." Interestingly, his surname sounded similar to that of Harry Houdini, one of the most famous illusionists all time. Many who watched the car show that day were convinced that this was Houdini's latest trick. Apparently, he was very angry for the mistake, he even visited Houdina and made him a tavern brawl.

Almost a hundred years have passed since then, and it is becoming increasingly clear that autonomous vehicles will change our lives, just as it was before with trains and planes. They will shape the future of road transport and can lead to a significant reduction in its costs. Finally, they can pave the way for new services and provide new ways to respond to the ever-increasing mobility needs of people and goods.

By 2021, autonomous taxis are to appear on the streets of London. Tokyo has been testing driverless taxis since August and wants to use such vehicles to drive visitors to the 2020 Olympics. The University of Michigan is already testing two autonomous buses that drive college students and staff around the campus to be widely available by 2025 and become the world's main mode of transportation in the future. According to the forecasts of experts from Allied Market Research, the global market of autonomous vehicles will be worth 54.2 billion dollars in 2019 and will grow to 556.67 billion dollars by 2026. This means an annual rate of as much as 39.47 percent.

In order to systematize the possibilities of autonomous vehicles, the organization SAE International (Society of Automotive Engineers) introduced their multi-level classification.

Marketing firm ABI estimates that 8 million autonomous vehicles (level 3 or higher) will be sold worldwide in 2025.

In 2017, approximately 94.5 million vehicles were sold worldwide. If car sales remained at a similar level for the coming years, in 2025 the share of autonomous vehicles in sales would amount to 8.5%.

Driving experience

Automated vehicles are not yet operational without human supervision. There are still many technical challenges that need to be addressed in order for the vehicle to be able to fully sense its surroundings, understand them and act accordingly, just like the driver does. However, there are already cars on the market that significantly support it (level 2 on the SAE scale).

When I got into one of them, I felt excitement, curiosity, but also anxiety. After all, I was about to entrust my life to the machine. But taking a semi-autonomous car to the streets of Warsaw, I could not resist the temptation to immediately try the latest technology, supported by numerous sensors, cameras and radar systems that allow me to navigate and avoid obstacles.

After pressing one of the buttons on the steering wheel, I start the driver assistant. The display in front of me shows that the car "sees" the lanes and cars in front of it and maintains a safe distance from them. He can recognize pedestrians, warn against a collision and brake when the need arises. At this point, I feel really redundant. The emotions increase significantly as the vehicle approaches the arc. There are a number of thoughts in my head. Will I crash into the lamppost? The mind tells you to take control, but despite my great fears, I rely on artificial intelligence. The car is maneuvering calmly, and a huge stone falls from my heart. The assistant holds the vehicle perfectly in the center of the lane.

While driving, I also get to know other possibilities of the vehicle. When I try to change lanes without turning on the indicator or when I slide a bit onto the lane next to it, the steering wheel starts to vibrate and the car immediately returns to its track. The car also recognizes road signs, including speed limits, so I don't have to worry about a possible fine. However, I can increase or decrease the speed at any time using a button on the steering wheel, and the car will adapt to my preferences.

Time for a break, you have to park somewhere. It was enough that I turned on the parking assistant and the car searched for and indicated free space. What's more, he parked himself, and I didn't have to worry that I would tear someone's mirror or scratch a bumper.

The assistant also works great in traffic jams or on the highway. Every few seconds, however, it reminds me that it is up to me to control the car. Then I have to put my hands on the steering wheel, otherwise the car will start to slow down until it stops. The assistant also does not work on all roads (it has problems, especially on dirt roads) and when it "loses" it, it suddenly turns off, which is warned by an audible signal. Another problem may be the need to constantly pay attention not only to the road and what is happening around us, but also to the operation of the machine. Therefore, with the current technology, it is difficult to judge what is more stressful: normal driving or autonomous driving. It certainly requires some refinement.

Therefore, the possibilities described above are only the beginning of what awaits us. Manufacturers would like drivers to have plenty of free time in the car. Instead of focusing on driving, they should spend more time working, playing or simply relaxing.

Security

But the key role and the greatest advantage of autonomous vehicles is to improve road safety, and in particular to reduce the number of fatal accidents, especially involving pedestrians. All by eliminating the weakest link in road traffic, which is man.

- Self-propelled cars will move harmoniously, and additionally, by communicating with each other, they will avoid dangers, while informing themselves about accidents, traffic jams, and even bad weather conditions, such as ice, snowstorms or downpours - he explains in an interview with TVN24 Magazine prof. Marcin Ślęzak, director of the Motor Transport Institute.

As he adds, thanks to the fact that autonomous vehicles will be "intelligent", they will know how to avoid dangerous events.

- Lack of human ailments such as mistakes, inattention, willingness to take risks, carelessness and physical barriers such as fatigue or lack of concentration will be completely alien to self-driving vehicles. The deficit of human defects, which are the cause of most road accidents, will automatically translate into improved road safety. Of course, the decrease in the number of accidents will go hand in hand with the increase in the share of self-propelled vehicles in the total population of cars on the road - he explains.

Hubert Fronczak (BMW Group Polska) points out that thanks to intelligent, connected systems, autonomous cars will be able to identify critical situations early. - They will also be able to reduce speed regardless of the existing restrictions and adapt the ride to the current road situation. Contrary to humans, automated cars will not be sensitive to distractions, will not get tired and thus will always consistently react, he emphasizes in an interview with TVN24 Magazine.

The European Commission estimates that thanks to autonomous vehicles it will be possible to achieve a very ambitious "vision zero", i.e. no fatalities on European roads by 2050.

Mobile sex tourism

However, there are many more benefits. Fewer accidents mean more savings, also for insurance companies. It is also a reduction in costly interventions by emergency services - the police, ambulance and fire brigades. The increasing share of autonomous vehicles on the market will also reduce fuel consumption and pollutant emissions, which is one of the main global challenges at the moment.

In addition, autonomous vehicles can enable mobility for people from areas with an insufficient public transport network (through car-sharing) or for people unable to drive independently (e.g. elderly or disabled people).

- Statistics show that out of nearly 4 million Poles who have a disability certificate, only about 200,000 have the right to drive vehicles. Even fewer have their own car. Let us remind you that fully autonomous vehicles will be devoid of a steering wheel and pedals, it is enough to communicate the destination to start the journey - says Ślęzak.

When asked whether millions of professional drivers may soon lose their jobs, he points out that in the development of autonomous vehicle technology, hopes are sought, not threats. - Looking only through the prism of the Polish transport market, where the phenomenon of the deficit of professional drivers is increasing, we should talk about self-propelled vehicles in the category of panacea. It is not by accident that advanced tests of autonomous trucks are carried out, which will be able to move in convoys consisting of several vehicles, only the first of which will be controlled by a human - says the director of ITS.

- It can be said that autonomous vehicles will certainly affect the broadly understood market of the Polish economy, including the labor market of, for example, taxi drivers, but for sure this revolution will bring more benefits than negative phenomena, even if they appear - he adds.

An interesting thesis was put forward by scientists from Great Britain - prof. Scott Cohen from the University of Surrey and Debbie Hopkins from the University of Oxford. They predict that autonomous cars will contribute to the development of ... sex tourism and the emergence of something similar to mobile "red light districts".

Who will be to blame?

Suppose an accident does occur. Who will be to blame then? - This is a good question that remains the subject of an expert debate, also taking place on a legal basis. Whether the traffic incident will be the responsibility of the driver (in the case of conditionally autonomous vehicles), or the system (software company), or the manufacturer of the vehicle or the manager of the "intelligent" road on which the vehicle traveled - remains an open question at the moment, but not without an answer - says Ślęzak.

It also indicates that another difficulty may be the need to update the software of such a vehicle and the responsibility for failure to do so.

- Nevertheless, we can agree that future legal acts regulating the admission to traffic of autonomous vehicles will contain appropriate provisions defining who is responsible for causing a road accident. The key to unraveling the blame will be the analysis of vehicle computers, for example by insurers. Hence the great emphasis on registering as many parameters as possible of the movement of such a vehicle, for example by using the so-called black boxes - Ślęzak points out.

Perhaps, however, the simplest solution, although it may require a surcharge, will be a form of a multilateral insurance policy that will relieve all parties from liability. So far, one person has been killed in accidents involving autonomous cars. The tragedy occurred in March this year in the suburbs of Phoenix in the US state of Arizona. Elaine Herzberg, 49, was crossing a four-lane road with her bicycle. According to the findings of the police, the woman entered the road outside the pedestrian crossing. At that point, it was hit by an Uber car, which was traveling at a speed of about 60 kilometers per hour in fully autonomous mode. Behind the wheel was, according to the regulations, a person in charge of the test drive and able to take control of the vehicle at any time.

Uber's internal investigation was to show that the sensors, cameras and radar systems designed to allow navigation (GPS) and avoid obstacles (LiDAR) were working properly. Unfortunately, the system that determines which objects around the car can be ignored has been poorly tuned.

American media suggests that Uber changed the software because it was so sensitive before that the car would brake even if sensors detected a wind-blown plastic bag flying in front of it. At that time, it was considered an exaggeration, because the ride becomes less smooth, uncomfortable for travelers and less safe. But the software was also supposed to malfunction after the changes were made. It did not recognize a woman passing in front of the car as an obstacle that needed to be braked.

Driving license not needed

The question is also whether you will need a driving license in the future? In the case of conditionally autonomous vehicles, i.e. equipped with steering devices, probably yes, if only because the driver needs to take control in a situation that is difficult to interpret by the autonomous vehicle's computer. But if the target autonomy level 5 is reached, the car will be a vehicle without any steering instruments. Consequently, the requirement to have a driving license no longer makes sense. Perhaps in the future it will not be possible to obtain a driving license at all.

In the case of the fifth and last generation autonomous vehicles, nothing prevents you from driving under the influence of alcohol in such a vehicle. However, experts point out that it will certainly be necessary to introduce systems monitoring passenger behavior to a greater extent than today, because a passenger behaving inappropriately can effectively disrupt the operation of many systems.

The trolley dilemma

With the development of autonomous cars, moral issues also arise. The question is, how should artificial intelligence behave in an emergency? Should an autonomous car kill the driver to save a group of passersby? What if there is both a child and an old lady on the road at the same time? Who should the vehicle hit then?

The situations described above are the classic trolley dilemma, also known as the crossover dilemma. This moral problem, considered in the 20th century, among others by the British philosopher Philippa Foot, it concerns a situation in which a person who has an influence on a vehicle has to decide whose life he should save at the expense of another (for example, a young or an elderly person).

In its basic version, it reads like this: "The cable car is out of control and is racing down the tracks. There are five people on its way tied to the tracks by a mad philosopher. one man is tied. What should you do? ".

Some of these dilemmas were to be solved by a study by American scientists from the Massachusetts Institute of Technology Media Lab. They wanted to establish priorities and rules for the behavior of autonomous cars. Its results, however, raise more questions than they answer.

Researchers created a platform called the Moral Machine, through which study participants could define their views on how autonomous cars should solve the trolley dilemma.

The response to the survey exceeded the wildest expectations of its creators. In the four years since the platform was launched, 40 million responses were obtained. Thanks to this large amount of data, scientists were able to analyze the differences in ethics depending on culture, economic situation and geographic location.

Moral Machine collected responses from respondents regarding nine cases: should an autonomous car save people or animals, passengers or pedestrians, more or fewer lives, women or men, older or younger, poorer or richer, law-abiding or violating the law, and above all, whether the car should take action or stay on its original course.

For example, one of the questions concerned a situation where a car could either continue driving, killing three older pedestrians in the process, or save pedestrians by hitting a barrier, which would result in the death of three young passengers.

It turned out that the answers were strongly related to the culture and economic situation of a given area. For example, respondents from countries such as China and Japan were less likely to kill older people in order to save younger ones. The researchers explain that this may be due to the cultural perception of the role of older people in Asian societies.

Another example is crossing the road in a place other than designated. In poorer countries, where the degree of the rule of law is lower, public acceptance for this type of violation is greater. Conversely, in economically developed areas, the tendency to sacrifice pedestrians who break the rules for those who pass on the lanes is more pronounced.

- Ethical dilemmas will not bypass artificial intelligence, but the "valuation of life" of other traffic participants will be as algorithmized, that is translated into the language of the machine, as programmers decide about it - says Ślęzak.

In the first generation of autonomous driving, there will be neither a technical possibility nor a legal framework for the car to make ethical decisions. - In the event of a dilemma, they will only be able to recognize whether the traffic area in front of them is "free and passable", whether it is "not clear" or "not passable". The car will be designed in such a way that in a critical situation it will immediately brake as much as possible. If the computer determines that the initiated braking maneuver is not enough to prevent a collision, it will look for the possibility of avoiding the obstacle and possibly change the direction of travel towards the free zone - says Fronczak.

What else needs to be done?

Research shows that the majority of European Union citizens have a positive view of autonomous cars, and 58% would like to try traveling with one of them. However, the European Commission points out that in order for automated mobility to gain public acceptance, it is necessary to maintain the highest safety and security standards.

"New threats must be faced, such as over-reliance on and misuse of technology. New issues must be addressed, such as the level of infrastructure support for autonomous vehicles and how this infrastructure and vehicles interact," reads the EC document of May 2018 .

Marcin Ślęzak emphasizes that from the point of view of digitization, the infrastructure that will ensure sufficient data transfer is important. - Only one autonomous car processes 1GB of data per day. The cars of the future can process much more information because they will exchange it with other vehicles and with the infrastructure. Therefore, we need a new, efficient teleinformation network in the 5G standard - he explains.

According to him, it is also necessary to improve the accuracy of the currently offered maps using GPS technology, and the current tolerance of around one meter should be minimized to a few centimeters.

- The accessories that are part of the entire driving autonomy system are also a challenge. Laser sensors and high-definition cameras will be miniaturized and their accuracy will be increased. In addition, their weight and energy consumption will be reduced so that they do not increase fuel consumption by the vehicle - he explains.

In his opinion, a reliable autonomous vehicle will have to communicate effectively with the road infrastructure and other vehicles, derive information from available digital maps, and in special cases, diagnose the environment itself using on-board sensors, so that the vehicle's movement is safe.

Hubert Fronczak also points to the challenge of technology. - Further development of sensory solutions, precise high-resolution maps, stable and fast information processing in the cloud, as well as artificial intelligence are necessary. Without machines that can learn, autonomous driving will not be possible, he points out.