Are self-driving cars really safe? We’ve been talking about the self-driving car for a while now. We even devoted a whole newsletter to the subject!
So what’s new?
Since we last wrote about self-driving cars, the test phases of manufacturers have multiplied and we are getting closer and closer to level 5, i.e.: the 100% driverless vehicle requiring no human driver. Nevertheless, the risks still exist. In March 2018 there was an accident in the United States. This was the first fatal accident involving a self-driving car. But could it have been avoided?
The accident took place on March 18, 2018 near Phoenix, Arizona. An UBER self-driving car ran over a pedestrian. Phoenix is known for its test drive area for driverless vehicles. Confidence in artificial intelligence has developed thanks to numerous tests that have been carried out, but this accident has put the brakes on UBER’s project, especially as the company is already going through an economically difficult period.
An investigation was opened to analyse the facts and the unfolding of the accident. According to the experts who were able to see the video of the dashboard, the radars installed on the self-driving car should have prevented the accident. The company Cortica, which develops artificial intelligence systems for driverless cars, also said the same thing. Moreover, the person in the car did not have time to react even if it was clearly established on the integrated video device that they did have their eyes on the road. Finally, the car didn’t slow down and hit the pedestrian at full speed. With all these factors in mind, where do you put the blame?
The latest information has revealed that the UBER autonomous car saw the presence of an obstacle (a person in this case) on the road but did not interpret it as an obstacle to avoid. The car had been programmed to consider certain obstacles as a “false positive” (such as, for example, the presence of a plastic bag on the road) and to not consider these types of obstacles as dangerous. UBER engineers had in fact lowered the sensitivity level of the car software to avoid the car stopping too frequently for “false-positives”.
What will happen next?
Most tests on self-driving cars show that they are a very safe. However, the majority of the tests are not carried out on public roads. The self-driving car still has a long way to go before it learns how to handle unusual situations. It also means that vehicles without drivers have to make complex decisions. Until then, engineers will have to work on how to adjust the integrated systems in these vehicles to find an infallible solution.