New York (AFP)
Uber's self-driving car involved in an accident that claimed the life of a woman in March 2018 had detected it but was not scheduled to recognize it as a pedestrian, said Tuesday the US transport regulator NTSB.
In a report issued before a hearing to determine the probable cause of the accident, on November 19, the agency claimed that Uber's technology "did not take into account the possibility of pedestrians crossing out of the nails".
The NTSB had already determined in a preliminary report that the software had spotted, nearly six seconds before the shock, the 49-year-old woman walking beside her night bike. But it is stated in the new document that "the system never classified it as pedestrian" but as an "object".
When the software determined that a collision was imminent, approximately one second and two hundredths before the impact, the emergency braking was deliberately not triggered because it would have caused a braking or a diversion of the trajectory of the vehicle too extreme . On the other hand, he sent an audible warning to the driver.
Uber had suspended all self-driving tests on public roads just after the fatal collision on a road in Tempe, a suburb of Phoenix, Arizona, before resuming them a few months later.
The company assured the NTSB that the new version of its technology would have correctly recognized the pedestrian and trigger controlled braking more than four seconds before impact.
The regulator's document also shows that Uber's test vehicles with its autonomous driving system had been involved in 37 accidents between September 2016 and March 2018. But Uber's car was the cause of the collision in only two cases.
The Tempe accident represented a snag in the development of autonomous driving by forcing most groups engaged in this technology to re-evaluate their safety systems.
© 2019 AFP