A recent report indicates that a GM Cruise autonomous vehicle made several technical errors in an incident last October wherein a pedestrian was dragged 20 feet after colliding with the AV. The report, commissioned by GM, was issued by Exponent, an independent, third-party engineering consultancy group. Cruise is facing heavy scrutiny following the incident, and now faces multiple investigations.
According to the report, the incident occurred at approximately 9:29 p.m. local time at the intersection around Cyril Magnin Street where it transitions to 5th Street and Market Street in San Francisco. The AV, nicknamed Panini, was stopped at a red light at the intersection alongside a human-operated Nissan Sentra, the latter of which was also stopped at the red light. Both vehicles proceeded across Market Street when the light turned green. A pedestrian then proceeded across the crosswalk against a Do Not Walk signal, at which time she was struck by the Nissan and thrown in front of the AV. The Cruise AV struck the pedestrian before coming to an initial stop in less than 0.1 seconds, after which the AV proceeded forward roughly 20 feet, dragging the pedestrian before stopping. The pedestrian suffered injuries, but survived.
According to Exponent’s report, the Cruise AV could not have predicted the accident. The AV also slowed down marginally just before striking the pedestrian. However, the onboard sensors failed to locate the pedestrian after the collision, as much of her body was out of view of the lidar detectors, and the left wheel rolled over the woman before stopping.
The onboard systems mistook the collision as occurring from the side, rather than the front, and the AV moved roughly 20 feet to pull over to the curb, dragging the pedestrian. This constituted another error, as the AV was already at the curb.
The report indicated that there were no issues with the sensors, but added that a human driver would likely have performed better in this particular scenario.
Cruise stated that it has updated its software to address the issues identified in the report. The autonomous vehicle division maintains that AVs can significantly reduce the number and severity of car collisions.