Fully autonomous vehicles may not arrive anytime soon

Published: July 31, 2018

CATEGORY:

Share:

People expecting fully autonomous vehicles to be on our roads soon may be left wanting.

Vehicles with self-driving technology implementations are already on the road and autonomous integrations are expected to increase over the next decade. We have heard a lot about how autonomous vehicles will transform road safety, auto insurance, and driver habits. It has also been widely predicted fully autonomous vehicles will be on our roads within 10-15 years.

Google, one of the key tech players behind the technology, has expressed its desire to fast forward the technology, eventually removing steering wheels and pedals from cars. Uber is already allowing autonomous tests.

In a blog post written by Russell Brandom of The Verge, manufactures predicting full autonomy in the near future are wrong. Such predications depend on the notion that the current technology will keep getting better. While this can be true for many tech services or products, there are no guarantees for autonomous cars.

The levels of autonomy in vehicles are as follows:

  • SAE Level 0 – human does everything, like current cars
  • SAE Level 1 – some in-car systems can aid the human in the operation of the vehicle
  • SAE Level 2 – the autonomous tech can complete some driving tasks, but human monitoring is needed.
  • SAE Level 3 – the system conducts some driving and monitors some of the environment, but human must be ready as backup
  • SAE Level 4 – the system can conduct driving tasks without any input from humans. However, the system only works under some conditions. This is where the current market is.
  • SAE Level 5 is when the car can perform all tasks without the need for a human driver.

“For a long time, researchers thought they could improve generalization skills with the right algorithms, but recent research has shown that conventional deep learning is even worse at generalizing than we thought,” Brandom wrote. “One study found that conventional deep learning systems have a hard time even generalizing across different frames of a video, labelling the same polar bear as a baboon, mongoose or weasel depending on minor shifts in the background.”