Uber the cause of the crash of a driverless car: the software found the victim but chose to ignore it.


NetEase news on May 8th news, according to TechCrunch reports, the net about car giant Uber unmanned car involved in a fatal car accident seems to be the software level. Specifically, the problem of deciding which objects need to be ignored and which objects need special attention is a problem, which puts the blame on Uber.

Uber the cause of the crash of a driverless car: the software found the victim but chose to ignore it.

Considering that the unmanned vehicle is equipped with a complex visual system and a backup system, either of them can help the car to perceive Elaine Hertzberger (Elaine Herzberg), who crosses the road directly from a vehicle laser radar and a front camera. However, the car didn’t even sign signs of braking, nor did it send out any sirens. Coupled with careless safety drivers, these failures led to Herzberg’s death.

The possibility of causing this tragedy includes: A. object recognition system failure, it may not classify Herzberg and her bicycle as pedestrians. This seems unlikely, because bicycles and people are the highest rate of system identification.

B: a vehicle’s high-level logic error, which determines what objects to pay attention to and how to deal with them. For example, it is found that bicycles on the roadside do not need to be slowed down. But suddenly turning on the driveway in front of the car requires a driverless car to take immediate action. This mimics human attention and decision-making and prevents cars from feeling lost when they detect new objects.

According to sources, Uber has identified B as the cause of the accident. Specifically, the system has set up problems, but those who should have paid more attention to it have been ignored. This means that Herzberg may have been discovered by car but is considered to be a false alarm. This is not good news, because unmanned vehicles have superhuman sensory capabilities, laser radar can stretch hundreds of meters in the dark, target recognition systems can track dozens of cars and pedestrians at the same time, radar and other systems can observe the surrounding roads.

But all of these feelings are like the human “brain”, the central processing unit, which gets information from cameras and other sensors and combines it with meaningful pictures around the world and makes it in real time to make a decision. This is the most difficult part of automobile manufacturing so far, as Uber has shown.

No matter how good your eyes are, if your brain doesn’t know what it is looking at or how to react correctly, it doesn’t make sense at all.

Uber issued the following statement, but did not comment on the above report: “we are actively cooperating with NTSB’s investigation. Due to respect for this process and our trust with NTSB, we can not comment on the details of the event. At the same time, we have started a comprehensive security review of our unmanned vehicle plans, and we have hired the former chairman of NTSB (Christopher Hart) to provide us with a proposal to build a comprehensive security culture, and our review from system security to the training of vehicle operators. “

Since this is a case without precedent, NTSB and other agencies may find it particularly difficult to report and the details are released very slowly. It is not normal for a company or individual to hide too much information before publication. (small)

AiShangHai LoveShangHai