Waymo has just released a large amount of data covering the 6.1 million miles of its cars since January 2019. This data contains detailed information about 47 collisions of Waymo cars during this period, which is a very rare lack of transparency in the industry.
What needs to be clear is that these collisions are a mixture of actual and simulated collisions, although Waymo said that it would not expect serious personal injury, and no serious personal injury. Waymo estimates that 18 of them were real and the other 27 were “simulated”, which means that they would indeed happen if it were not for the intervention of the safety driver.
Let us look at some of them. In the first example, the Waymo car was hit, not Waymo crashed:
The two actual collisions involved the Waymo vehicle being hit on the rear bumper while driving straight at a constant speed at or below the speed limit. In a collision, the Waymo driver decelerated to a constant speed while driving on the speed bump. In another collision (Figure 4, Event C), a Waymo vehicle traveling at the speed limit was hit by a vehicle that exceeded the prescribed speed limit of 23 mph. The severity of both collisions was S1. In the second collision in this group, the airbag deployment occurred in the crash.
The second is possible road rage, a side note, when it points to self-driving cars, it becomes even stranger.
A single simulated event in this group (line 17 in Table 1) involved a vehicle that turned to the lane in front of Waymo and braked immediately after cutting in, despite the absence of any obstacles ahead (consistent with the constraint of motivation). The Waymo Driver was simulated to achieve full braking in response to the braking of another car, but after simulation, it came into contact with the lead car at a relative impact velocity of 1 mph (severity S0).
Finally, the worst simulation was performed:
The simulated collision (event H) in Figure 9 depicts the vehicle turning left on the path of the Waymo vehicle. The other car has no right of way at any time leading to the described sequence of events. The simulated response of Waymo drivers to vehicle actions is to start braking before entering the intersection. The simulated full braking was achieved, which reduced the speed by 12 mph before the simulated collision. Based on the vehicle mass at the time of the collision and the simulated vehicle speed and geometry, this event is classified as S1, and the airbag is expected to deploy. It is the most severe collision (simulated collision or actual collision) in the data set, close to the boundary between the S1 and S2 classifications.
The transparency of the report is an important reason why experts say Waymo is leading the autonomous driving race, although you might think of Tesla’s ideas. That’s because, compared with Tesla, Waymo is actually inviting review, and the product Tesla sells is called “fully autonomous driving”, which is not fully autonomous at all.
You can read the full Waymo report Here, Herewith Here.