Uber dashcam recording of the crash
Based on the initial findings that experts in the field of AV technology have gleaned by viewing the video of the accident in Arizona, suggests that there was a "catastrophic failure" of performance. Taking into account human reactions when faced with such a situation, the AV technology did not even notice the woman, did not detect her presence and therefore did not react at all. Basically, based on the video recording the experts claim that Uber's LiDAR technology failed.
According to Professor David King of Arizona State University, a specialist in transportation planning said that "This is exactly the type of situation that Lidar and radar are supposed to pick up, this is a catastrophic failure that happened with Uber's technology."
What is also noticeable is that the safety driver in the car was not watching the road. Otherwise, he would have hit the brakes immediately.
Duke University Professor Missy Cummings stated that "This safety driver was not doing any safety monitoring. The problem of complacent safety drivers is going to be a problem for every company."
Video
According to another expert, Professor Bryant Walker Smith, of the University of South Carolina law school, he suggested that the nearly 2-second interval when the pedestrian became visible was ample time for an immediate reaction, "This is similar to the average reaction time for a driver. That means an alert driver may have at least attempted to swerve or brake."
The Uber AV was traveling at a speed of 38 mph when it crashed at 10pm, and according to local police, Uber was not to blame.
My opinion: There are a lot of "experts" out there, professors and such like that sit in the ivory towers and pass judgment on everyone without having any field time or real expertise in the area they are criticizing. All these "experts" are just detractors of Uber, taken on to find fault in Uber and were most probably financed or backed by Uber haters.
The only real experts in accidents and crime scenes are the police investigators that have been doing their job for years, are on the scene collecting real evidence, and relying not just on a video, but on testimonials, physical evidence as well as viewing the video.
Having stated this, there is a question for the LiDAR experts, while a human driver might not have picked up the shadow or form of the pedestrian, should not the lasers and other scientific methods have picked out the form even in the dark? Lasers do not need light to see; they are light.
Here is a description of what a LiDAR does. Courtesy:
The LiDAR gun clocks the time it takes a burst of infrared light to reach a car, bounce off and return back to the starting point. By multiplying this time by the speed of light, the LiDAR system determines how far away the object is. Unlike traditional police radar, LiDAR does not measure the change in wave frequency. Instead, it sends out many infrared laser bursts in a short period of time to collect multiple distances. By comparing these different distance samples, the system can calculate how fast the car is moving. These guns may take several hundred samples in less than half a second, so they are extremely accurate.
This means that the LiDAR should have received a reading back from the circumference of its surroundings and read that there was an object approaching from the left front side. It did not, which raises three questions:
- Do LiDAR work in the dark?
- Does a LiDAR need infra-red or night time vision to enhance its performance in the dark?
- How fast before a reading triggers a response?
As we saw from the video, the safety driver realized there was a person in the road only after she became visible in the front light, and he did not react in time to brake, which is a natural human reaction. However, even if the driver was driving and not the AV, would it have made a difference?
These are the questions that need to be asked and will be asked by the local Police, and Federal transportation authority's when investigating this accident.