Chances and challenges of autonomous driving
Next to electromobility, autonomous driving is without doubt the hot topic in the automotive industry. The question is by now no longer if, but when, autonomous driving will become our reality. The competition of the car manufacturers and their respective development partners is in full swing and the IT industry is working flat out on the digital infrastructure. However, transition to the new kind of mobility is tedious. Even though developments and patent applications are advancing at a furious pace, the way onto the road and the legal basis have as yet not been created. Apart from a whole lot of chances, there exist also a whole lot of challenges, ethical crunch points and on top of that different basic conditions worldwide. A study of the auditing and consulting company KPMG showed that the Netherlands, the USA and Singapore offer the best preconditions for driverless vehicles. The comparison was made in the areas of policy/legislation, technology/innovation, infrastructure and customer acceptance. Although roughly half of the relevant patents are held in Germany, German car manufacturers are rather behind when it comes to developing and testing the technologies, at least in their own country. This is not least due to the liberal legislation in the USA or even in China. Thus, highly automated driving would have been possible with the Audi A8 since 2017, but the software required for this may not yet be legally activated. However, apart from legislation, the development backlog in Germany and in the whole of Europe is surely also due to the “Safety First Approach”, which is by all means confirmed by accident reports from the USA involving autonomous vehicles.
The five levels of autonomous driving
Usually, five levels of autonomous driving are referred to, which range from assisted driving to the driverless vehicle. In Europe, we are currently predominantly dwelling around level 3, while Waymo with its “Early Rider Program” already launches driverless taxis in the sparsely populated suburbs of Phoenix, Arizona. In China, the start-up company WeRide in Guangzhou most recently started driverless test drives. Elon Musk recently announced that at Tesla, the basic functionality for level 5 is to be reached before the end of the year.
Level 1 -> This stage is referred to as assisted driving, which, for example, already includes cruise control. The automatic lane keep assist system also forms part of level 1 autonomous driving.
Level 2 -> During partially automated driving, the driver must keep a constant eye on the traffic and always be in control of his vehicle. However, in some places, for example on the highway, the vehicle can autonomously keep in lane and accelerate and brake. In addition, level 2 includes the overtaking assist system and the parking assist system, which allows the driver to take his hands off the wheel.
Level 3 -> In many places, we are currently on the verge of level 3 of autonomous driving - highly automated driving. At this stage of development, a vehicle is able to drive autonomously in certain scenarios, without the driver having to keep his hands on the wheel and his eyes on the road. However, at a signal emitted by the vehicle, the driver must be able to fully take control on short notice. Thus, the driver could, for example, be reading a book for a limited period while driving on the highway under regular conditions.
Level 4 -> With fully automated driving, the vehicle masters longer distances completely independently and the driver turns into a passenger. During fully automatic driving, the vehicle driver can, for example, sleep and is not liable for traffic violations. The system must be designed in such a way that the vehicle can be switched to a safe state in situations that exceed its limits, even if the occupant cannot or does not wish to intervene.
Level 5 -> In fully autonomous driving, there will only be passengers. At this level, the vehicle can cope with any kind of traffic situation, such as independently driving through a roundabout or crossing an intersection In case of an accident, the manufacturer would be held liable; at this level, passengers would not be liable. However, there is not yet a legal and actuarial basis for this.
Chances offered by the new mobility
Apart from the additional convenience autonomous vehicles can offer to drivers and passengers, there are numerous other profitable aspects. Compared to human drivers, the systems never become tired, unfocused, distracted or even aggressive. Reaction times are much shorter and it is safe to assume that there will be an immense gain in safety on the roads. The number of traffic deaths could partly be reduced by up to 25 percent. While driving, occupants could use the time for other activities, such as working. Even persons who due to their health, age or other circumstances are normally not able to drive a vehicle could become mobile again. Or we could think of the automated transport of goods and the reduction of CO2 emissions through optimized traffic routing and passenger transport.
The vehicle of the future will be networked with numerous mobile devices in the vicinity, register the cyclist already from a great distance by way of his smartphone or warn following vehicles of obstacles. Autonomous vehicles will be able to find a parking space in the hustle and bustle of cities by means of overarching platforms, which also integrate the parking guidance systems, without the occupants even having to be in the vehicle. The list of advantages we can hope to gain through autonomous driving could be continued and some of you may find yourself already dreaming of KITT from the Knight Rider series.
About the challenges or: Algorithms and ethics
From a technical point of view, most of the scenarios described have long since been feasible. Theoretically, numerous complex situations in road traffic could already be tackled by means of AI. However, there are still large hurdles. It is therefore to be expected that in the forthcoming transition period, autonomous vehicles will continue to drive alongside non-autonomous vehicles in road traffic for several years, if not decades. Human drivers, but of course also cyclists and pedestrians, do not always act in accordance with the rules and with that could quickly push the intelligent systems to their limits. On the other hand, human drivers could be dumbfounded by an autonomous vehicle that very strictly adheres to a solid line when approaching an obstacle and therefore makes an unforeseen stop.
Autonomous driving also raises ethical questions: According to which maxims should algorithms be formulated when exceptional situations arise in which people are harmed? The trolley problem is often cited in these discussions, as is the problem of weighing and of utilitarianism: If for instance, a situation presents itself where a person in the street can only be evaded by endangering the occupant himself - how should the system then decide? Is the system in this case allowed to offset the number of human lives that would be in danger? Or even base its decision on age or other physical features? The “moral machine” experiment even goes as far as to illustrate how persons from 233 countries would solve the trolley problem and where the majority of them would rather have a criminal run over than a dog. In Germany, the situation has been unambiguously clarified by the Ethics Committee: Qualification according to personal features and offsetting of victims is strictly prohibited.
The next hurdle in advancing autonomous driving is the question of liability in case of accidents. Who is to blame if the technology fails? The automation level plays a decisive role in this regard. Up to level 3, the driver is undoubtedly responsible, but anything beyond this level of automation is not quite as clear-cut. If the occupant is only a passenger, the manufacturer must be liable. At levels 3 and 4, the question arises as to how exactly restoration of the perceptive faculty must be regulated, i.e. how quickly the driver must be able to regain control of the vehicle. At the latest at level 4 it would have to be demonstrated by means of a black box whether the vehicle or the driver was in control at the time of the accident.
In January 2021, the United Nations adopted binding international rules for automated driving at level 3, which provide clarity at least for roads where pedestrians and cyclists are not permitted. The main issue here is how the transfer of control between the driver and the system must take place. However, regulations on cyber security, liability and technical specifications are also included.
Of course, data security is also a major issue when it comes to the challenges of autonomous driving. It can be assumed that a networked car generates around ten terabytes a day. To avoid data chaos, all networked units must function and communicate with each other at a superordinate platform level. This is where concepts such as the monetization of data and cybersecurity come into play. And when we talk about obstacles in the area of these data volumes: of course there is also the GDPR.
Another hurdle is the partly still extremely high priced technology which so far has made an introduction of autonomous driving into the private customer sector inconceivable. In particular technologies in the Lidar sector are expensive. This is probably one of the reasons for Tesla's lead in the race: Tesla initially only passes on the costs of the expensive technology to the customer to a limited extent. The actual full digital equipment must be activated at a later stage at a charge once all external requirements have been met. Higher automated driving could be run as a kind of subscription not only at Tesla. This type of sales strategy is already known from the BMW updates, where software packages permit, for example, an enhanced language assistant or an improved brake intervention against payment (see Sueddeutsche Zeitung, article from June 19, 2020).
Components which spur on autonomous driving
But not only the car manufacturers are preoccupied with the new mobility. The suppliers are also on board - Bosch ranks first in terms of the number of patents in the field of autonomous driving.
MinebeaMitsumi too is developing a multitude of products and components for autonomous vehicles. For example, resonant devices can assist in warning drivers of level 2 and 3 vehicles when there is a need for intervention in dangerous situations. In the Lidar sector, the FDB motor with extremely high running smoothness and running precision is often used. Electromagnetic compatibility will play an ever increasing role not only in autonomous driving: In this field, coils are used for optimization purposes. There are numerous new areas of application for the different DC and BLDC engines as well as for the MinebeaMitsumi actuators, for example automatic tailgates and electric parking brakes. Break-by-wire applications are also conceivable. Stepper motors are used in the head-up display. Battery sensors from MinebeaMitsumi make it possible for autonomous vehicles to know on their own when they need to recharge.
But also backlights, camera connectors (e.g. for driver monitoring, seat occupancy detection and mirror digitization) and antennas for car-to-car communication as well as resolvers can be used in the autonomous vehicle and make their contribution to tomorrow's mobility.
The MinebeaMitsumi subsidiary Paradox Engineering approaches the trend topic from the promising Smart City perspective: Smart Lighting or Smart Parking network nodes, on top of their specific services, would also form virtual corridors for autonomous vehicles and allow safe and robust navigation and traffic control, coordinates, navigation routes, etc. They would also help by providing reference regulations for the area, situation awareness based on sensing of the area for improved security, information about other vehicles in the area. Nodes by Paradox Engineering are Blockchain enabled, ensuring maximum security in the exchange of data and information and therefore safety for users and citizens alike.
This approach was originally developed and patented for drones to define flight paths for autonomous parcel deliveries or surveillance services. However, looking at it from a broader point of view, the concept can be applied to any unmanned, autonomously moving object and offers a valuable approach to autonomous driving: the technology is ready.