• Home
  • traffic
  • The car of the future: Why we won't be driving fully autonomously in 2030

The car of the future: Why we won't be driving fully autonomously in 2030

Car of the future
Subscriptions & booklets

K aum the hype about electric mobility has calmed down, the car manufacturers drive their studies on the topic of autonomous driving through the village. They suggest that in 15 years' time we will not be drivers, but passengers, who will turn their seats at will, turn to fellow passengers, read or sleep. 15 years are two model generations and therefore little time in view of the problems that have to be solved.

Two model generations ago the first radar-based distance cruise control came onto the market, which today works very well, but by no means perfectly, like the one Regularly prove test cars in the auto motor und sport fleet regardless of their brand. In curves, guardrails are mistaken for cars driving ahead (and warned violently) or vehicles cutting in to the side are only registered late. And should we turn our backs entirely on the road in two model generations? Have fun.

For their autonomous demo drives, the manufacturers have to cope with a reality that is not yet suitable for the very big goals: High-resolution map material is missing, and since the accuracy of GPS systems is by no means sufficient, there is a differential -GPS used. Additional transmitters set up on the roadside create an artificial precision. This was the only way for the Audi RS7 to do its fast laps in Hockenheim at the 2014 DTM finale. Thanks to differential GPS, the Mercedes F 015 also follows a pre-programmed route. Since it has no sensors, nobody is allowed to come too close.

Authorization through increased safety

Mercedes refers to successful campaigns such as the Bertha Benz drive, in which an S-Class drove independently from Mannheim to Pforzheim. Trips like those organized by Audi, BMW and many other suppliers are well-rehearsed (and quite impressive) tricks, for which every meter of distance has to be measured and recorded in order to be recognized by the sensors.

So Developers report that special construction teams supplement missing lane lines in advance, as otherwise no lane guidance would be possible. But even if the important roads should be recorded in a few years, problems will remain for which no solutions are currently in sight.

Example of road conditions: The coefficient of friction between tires and road determines how long braking distances are or how fast a car through a curvecan drive. There are already big differences between dry and wet roads, with ice the drop in the coefficient of friction is enormous. However, there are currently no sensors that can reliably predict the coefficient of friction. An oil trail, for example, can only be recognized by the minimal color differences on the road surface, which an experienced driver can perceive intuitively in order to choose a different line at the last moment. The robot car falls into the trap here. Car companies like to bring carto-car technology into play, in which cars ahead warn their successors. But one is always the first to encounter an imponderable.

The sensor developers also still have a lot of work to do: If the autonomous car is no longer able to continue, it must return the steering wheel to the driver within a certain period of time give. Most manufacturers put this transfer time at around ten seconds. It takes a person that long to put his newspaper aside, to see the traffic and to grip the steering wheel - a seat rotation is not even planned.

At 130 km /h a car puts in ten Seconds already 360 meters back, but the best radar sensors only manage 200 meters, not to mention cameras. In addition, radar sensors react poorly to ice, even heavy rain brings them out of step. In the city, computer brains are completely overwhelmed, as the behavior of pedestrians at the edge of the road cannot be predicted.

In order for robot cars to be approved, they also demonstrably have to cause fewer accidents than humans. And less than people plus assistance systems. But radar and Co. are already making cars safer. Statistically, a serious accident happens every 7.5 million kilometers. In order to prove statistically relevant that autonomous cars drive safer, their developers would have to travel billions of kilometers.

The legal situation is also problematic: If an autonomous car is to blame for an accident, the developer goes to jail according to current law . Who would like to take on this responsibility? The problems have been recognized in Germany and other countries, but in view of almost 30 EU countries and 50 US states, a uniform legal framework is still a long way off A situation arise in which an accident can no longer be prevented, but the opponent can be selected: cyclist on the left or child on the right? It is simply forbidden to make decisions that favor a life. However, such a decision must be programmed into an autonomous car. Or should a random generator decide?

1. Incalculable grip

Dry asphalt, wet, oil or even ice? The coefficient of friction determines how long braking distances are and how fast you can drive through corners.At present, however, there are no sensors that can reliably predict the coefficient of friction.

2. Limits of radar

The range of radar sensors is far too short, and raindrops can be enough to throw them out of sync. In addition, cameras do not recognize road markings when the sun is low, without which nothing works.

3. Safety difficult to measure

For the legislator to allow autonomous cars, they have to cause fewer accidents than their conventional colleagues. However, they are also becoming increasingly safer, as the sensors of the robot cars are already being used in assistance systems.

4. Complicated liability issues

According to current law, developers are liable for accidents caused by an autonomous car - in extreme cases up to jail terms. Regulations of 50 US states and almost 30 EU countries stand in the way of a uniform legal framework.

5. Ethical conflicts

Robots keep a cool head even immediately before a crash and will therefore be able to choose the opponent in the accident. So they have to make decisions that are forbidden: small car or SUV? Child or oncoming traffic?

Insufficient forecasts for the time after 2020

Respect for all the vigor! However, given the current state of technology, it is unlikely that we will drive fully autonomously in 15 years. The time horizon for the specific introduction of automated driving functions ends from 2018 to 2020 for a reason - for almost all manufacturers. Until then, ACC and lane keepers should merge into a motorway assistant, which we leave the wheel for a certain period of time at higher speeds. A few parking tricks should also be added. For the period after 2020, however, there are only nebulous forecasts and full-bodied announcements, against which the otherwise reluctant German manufacturers have to be measured. If you want to be driven through bustling city centers and sit opposite your passengers, you will still have to take the S-Bahn in 2030.


Leave a reply

Name *