Former Google Waymo self-driving car is a simulation result with 'accident avoidance ability that surpasses humans'



On March 8, 2021, Waymo, a self-driving car development company under the umbrella of Google's parent company Alphabet, announced the results of an experiment simulating an accident that occurred in a traffic accident-prone area in the United States. Among them, Waymo reports that the company's autonomous driving technology '

Waymo Driver ' was able to avoid accidents with a high probability even when a human driver causes a traffic accident.

Waypoint --The official Waymo blog: Replaying real life: how the Waymo Driver avoids fatal human crashes
https://blog.waymo.com/2021/03/replaying-real-life.html

According to Waymo's director of safety research, Trent Victor, while much research has been done on the safety of self-driving cars, 'what are the consequences of self-driving cars in situations where humans cause accidents?' No research has been done on 'Will you issue it?' Therefore, Victor et al. Conducted a simulation using the results of Waymo's public road driving test covering 20 million miles (about 32,186,880 km) and the data of statistical surveys on actual traffic accidents.

Waymo's research team first extracted accident investigation records from Maricopa County, Arizona, the region with the highest pedestrian fatalities in the United States, from data released by the US Department of Transportation's Road Traffic Safety Authority. We have collected information on the fatal collisions that occurred between 2008 and 2017 in the county city of Chandler.

Then, 52 patterns of simulations were performed assuming that 'the driver of the car on the side of the perpetrator who caused the traffic accident was the Waymo Driver'. As a result, the Waymo Driver's 'probability of avoiding a collision in an accident' has reached 100%.



For example, the following is a simulation result of an accident in which a car that entered an intersection from the bottom of the screen with a speed violation and ignored the signal collided with a car on the victim's side that entered the intersection at a green light from the right of the screen. If you replace the driver of the car that caused the accident with the Waymo Driver, you will not enter the intersection at the red light and you will not cause a speed violation, so it is considered that an accident cannot occur. If you click the thumbnail below, you can see the simulation in GIF animation.



Also, in the same situation, the victim's car was replaced with Waymo Driver as follows. The Waymo Driver accurately recognized the movement of the car that caused the accident and rushed into the intersection at a tremendous speed, ignoring the signal, and slowed down in advance, so it was possible to avoid a collision.



Simulations assuming that the victim's car was a Waymo Driver showed that in 82% of cases the collision was completely avoided and in 10% the scale of the accident was reduced. On the other hand, in 8% of all cases, the scale of the accident did not change, but Victor said, 'All cases crashed into the rear of the Waymo Driver, which was stopped or running at a constant speed. Because of the pattern that was done, the Waymo Driver had little chance of avoiding an accident. '

'94% of accidents are due to human error, so we believe that replacing human drivers with Waymo Drivers can improve traffic safety,' Victor said in a report. He concluded that the results of this study supported his idea.

in Ride, Posted by log1l_ks