Uber's autopilot collision case, which has attracted widespread attention, showed that the autopilot's induction system recognized the deceased, but the software "decided" not to respond.
On May 8th local time, the US media The InformaTIon quoted two informed sources as saying that the American car-hailing software company Uber Technology (Uber, hereinafter referred to as Uber) determined that the recent death of an Uber self-driving car may be Caused by software failure: The vehicle sensor successfully identified the victim before the accident, but the software system "decided that there is no need to react immediately."
However, this is an internal investigation conducted unilaterally by Uber. The joint investigation between the National Security Transportation Board (NaTIonal TransportaTIon Safety Board, NTSB) and Uber is still ongoing, and the official preliminary investigation results will be announced in the next few weeks.
U.S. National Transportation Safety Board member inspects accidental unmanned vehicle damage
Software inner OS: In my eyes, you may be a plastic bagThe Uber self-driving car crashed to pedestrians in Arizona in March caused another big discussion about the safety of autonomous driving. Even more disturbing than the accident itself, Uber ’s findings proved that something went wrong with the software system of autonomous vehicles.
Uber's self-driving car crashed into pedestrians before the accident occurred
Like many self-driving car systems, Uber has set its software system to automatically recognize error messages, including objects on the road that will not affect driving, such as plastic bags passing over the road. Uber executives admit that this setting will cause the vehicle's delayed response to road objects.
The working principle of Uber's first unmanned vehicle: the high-top LiDAR can emit 1.4 million laser signals per second to draw a 3D map of the surroundings of the vehicle. The color camera can color the 3D map, so that the vehicle can recognize the change of the signal light. The front, rear and surrounding LiDAR modules of the vehicle can quickly identify obstacles in the blind spot. 20 cameras can identify braking vehicles, pedestrians and other obstacles. The roof antenna can locate itself through GPS. The cooling system ensures that the vehicle maintains a suitable temperature.
A person familiar with the matter mentioned at the beginning of this article said that the software's automatic error recognition feature setting was overdone, so that the vehicle's response was not fast enough. It can be understood in this way that in Uber's self-driving cars, the "authority" granted by humans to the software is too large, but the software does not have the ability to make judgments that match the "authority".
At the moment of the accident, Uber's software may really identify the victim who pushed the bicycle across the road as a plastic bag.
The front image of Uber lidar system
Safer or more comfortableSo the question is coming, why should Uber give the software "too big" automatic recognition permission?
The sole purpose is to make passengers more comfortable.
It is conceivable that if self-driving cars detect small stones, pieces of paper and other objects that do not affect the driving at all, they should keep a distance or make emergency braking to ensure safety. I am afraid that the people on the car will be stunned all the way. Therefore, improving comfort as much as possible is a direct reason for Uber to let the software automatically recognize it.
In contrast, Google ’s parent company Alphabet ’s self-driving car company Waymo and General Motors ’self-driving division Cruise AutomaTIon are not so focused on comfort.
According to The Information, people who took the vehicles of these two companies said, "It feels very bad and often brakes abruptly. Even if there is no object in front of the driving, the vehicle will be recognized as a" moving object "passing by, or there is a danger in front. Therefore, emergency braking from time to time. Obviously, Waymo and Cruise Automation have not discounted safety to improve comfort.
Comparison of safety warning sensors in Volvo XC90 and Ford Fusion. The number of safety warning sensors in Uber's unmanned car Volvo XC90 is less than that of Ford Fusion. The car was used for Uber unmanned vehicle testing. Uber gave up using this car since last year. Ford Fusion configuration includes, vehicle side and rear camera, roof laser sensor, front camera, vehicle front, two wings and rear laser sensor. The Volvo XC90 configuration includes a 360-degree coverage radar, a front camera, a roof laser sensor that can scan 360-degree measurements of the surrounding environment, and side and rear cameras of the vehicle.
How to find a balance between safety and comfort is not easy.
According to US media reports, in order to ensure safety, the safety staff of Uber's self-driving car had to take over the vehicle every mile (about 1.6 kilometers) during the road test last February. According to a New York Times report, until March this year, before the fatal accident, Uber could not achieve an average of 13 miles (about 21 kilometers) completely unmanned.
In this accident, should the safety officer who took over the vehicle in time when seeing the obstacle (or person) take measures when the accident occurred? How much responsibility do security officers take on this matter? It is not known until the official investigation is released. But Uber said in a previously published book that the employee is still in Uber.
Real-time recording of images of safety officers and victims in the car before the accident
Not the first to eat crabs, but the first to be bitten by crabsUber is not the first in autonomous driving road tests. In recent years, there have been many traffic accidents involving autonomous vehicles. However, Uber is the first one to be killed by automatic driving.
After the death of Uber's autonomous driving accident, Arizona banned it from testing autonomous vehicles in the state, and Uber suspended the road test for autonomous vehicles in California. Subsequently, the US chip maker Nvidia announced that it will suspend autonomous vehicle testing worldwide. Toyota also said it will temporarily suspend self-driving car testing projects in the United States.
However, car companies such as Volkswagen, BMW and PSA said that they will not change the autonomous driving program. Baidu CEO Li Yanhong also made it clear that the progress of Baidu's autonomous driving project is not affected by the Uber incident.
Of course, the development of autonomous driving technology will not be stalled by a fatal accident, but the software's "decision does not need to be done" investigation results prove that authorizing software to make life-threatening decisions requires more conditions.
The DC Source System has high power density, short response time, high accuracy and full protection. The programmable power supply has been widely adopted in production line, burn-in test, laboratories and integrated system etc. With various communication interfaces, apm DC Power Supply can be remote controlled from computer.
Some features as below:
- With accurate voltage and current measurement capability.
- Coded Knob, multifunctional keyboard.
- Standard RS232/RS485/USB/LAN communication interfaces, GPIB is optional.
- Remote sensing line voltage drop compensation.
- Equips with LIST waveform editing function.
- Use the Standard Commands for Programmable Instrumentation(SCPI) communication protocol.
- Have obtained CE certification.
120V DC Source System,Industrial Power Supply,DC Solar Small Power System,Dc Power Source For Lab Test
APM Technologies Ltd , https://www.apmpowersupply.com