Can Autonomous Cars Get Around The Corner?

It may sound simple, but it’s not so easy for machines.


With all the buzz about autonomous vehicles, one big problem still remains. People and autonomous vehicles don’t get along well together.

Cars are predictable. People are not. Cars behave according to a predetermined set of rules. People look for opportunities and take them, even if they’re sometimes not the wisest choice, such as passing around a blind curve or driving too fast on slippery roads.

This is the whole argument in favor of autonomous cars. If human drivers are taken out of the picture, there will be fewer accidents because machines respond faster, more accurately and more predictably in the vast majority of cases. So even if autonomous vehicles are following each other very closely, they can respond quickly enough to avoid hitting the car in front of them. And they never hit cars while backing into a parking spot.

Still, there are a couple of problems that continue to vex carmakers. One is that cars don’t always recognize objects correctly. This is the thinking behind new industry regulations calling for autonomous vehicles to be able to identify a child at all times.

“Intervention strategies aimed at protecting children recognize that children are not little adults,” according to the ISO/IEC guidelines. “Children’s susceptibility to injury and the nature of their injuries differ from those of adults. Such intervention strategies ideally also consider reasonably foreseeable use of products or surroundings. Children interact with them in ways that reflect characteristics of child behaviour, which will vary according to the child’s age and level of development. Intervention strategies intended to protect children therefore often differ from those intended to protect adults.”

In other words, children are less predictable than adults when it comes to traffic rules. Cars are supposed to stop for them. But there are many ethical discussions underway about who or what a car should hit if an accident is unavoidable and why. So far, there are no good answers.

A second problem is that while autonomous vehicles can be programmed to obey stoplights and traffic signs, they cannot turn right through a throng of pedestrians on a crowded city street. For anyone who has driven in a crowded metropolitan area, this isn’t always predictable. Cars typically nose their way into an opening, and then dart around the corner. But they also may have to pull forward partially and wait until after the light turns red before they can complete the turn. A human driver might have to wait 20 seconds to make the turn, but an autonomous vehicle could be stuck there for hours because it doesn’t nose forward. The only way to solve that is to have turn signals on every corner where no people are allowed to cross.

This problem was first identified in the poultry industry with “chick sexing,” which is a well-known issue in artificial intelligence labs. Large commercial hatcheries sort young chickens according to their sex, and workers generally get the hang of identifying males and females after a couple weeks. Machines, on the other hand, have never been very successful at this.

Can you identify the males? Source: U.S. Department of Agriculture

Whether this changes in the future remains to be seen. Artificial intelligence research is getting an infusion of capital from a variety of industries, particularly automotive, and research is underway around the globe. But will that help autonomous vehicles turn the corner? Or will the automotive infrastructure—roads, bridges, traffic lights, street signs, education—all need to be rethought to make way for vehicles without drivers.

These are issues that are only starting to be understood, and they depend heavily on whether machines can go beyond probabilities to make judgment calls that may not always be correct.

Related Stories
Reshaping Automotive Design
Conflicting goals, evolving standards and the need for new methods and tools make this an interesting market for chipmakers.
Auto Chip Test Issues Grow
Semiconductors used in cars have higher quality and reliability requirements than most chips, but they have the same cost and time-to-market pressures.
Faster Commoditization In Cars
As automakers rev up autonomous driving, cost cutting needs to happen at an unprecedented pace and scale—even for new technologies.
China’s Ambitious Automotive Plans
After years of playing catch-up to multinationals, the country is targeting an opening with assisted and autonomous vehicles.