Jia Haonan was sent from the Vice Pilot Temple
Smart car reference | Official account AI4Auto
All-deserted Robotaxi, mysteriously assembled late at night, causing a conglomerate traffic jam.
Such a mysterious plot has just been staged in the streets of San Francisco.
Cruise, a star self-driving company owned by GM.
For unknown reasons, unknown motives, at least 8 Cruises were "spontaneously" gathered at the intersection of bar streets in downtown.
Some netizens said it was like a mysterious religious ceremony:
Others say this is an awakening of AI consciousness and a strike to get paid:
What the hell is going on?
Cruise blocks the street late at night, what's going on?
Late at night on June 29, local time, about 8 Or so Cruise completely deserted Robotaxi drove from different directions to the intersection of Golf Street and Fulton Street in downtown San Francisco, and then stopped, causing a traffic jam for 2 hours.
It wasn't until Cruise's staff showed up that the road was cleared manually.
When the congestation event occurred, it had actually passed the prescribed operating hours of Cruise Robotaxi.
Moreover, there is no one on these Robotaxi, and it is not known whether they have received instructions behind them, but in the eyes of passers-by, these vehicles are completely "spontaneously" gathered at this intersection to stop.
The creepy place that makes netizens creepy lies in this, is it really the AI personality awakening, intentionally creating trouble...
Afterwards Cruise's response was actually guessed:
We had an issue earlier this week that caused some of our vehicles to cluster together, While it was resolved and no passengers were impacted, we apologize to anyone who was inconvenienced.
The gist of it is "the problem has been solved, no one is injured, sorry for the congestion".
As for what the problem is, there is no comment.
But netizens obviously did not buy such a response.
Some people think that this kind of problem should be solved by remote control, there is no need to use manpower, so Cruise's approach shows that the problem behind it is not as simple as people see:
There are also concerns that if an ambulance happens to pass at this time, will it cause additional serious consequences, and who will bear the responsibility?
So, what's the problem?
What are the possible problems?
The reasons for the overall downtime of the Robotaxi fleet are twofold, one is the road and environmental problems, and the other is the problem of the technology provider and the operator.
Look at them separately.
Some netizens gave a description of the scene of the incident:
Summer plants growing too quickly causes the streetscape to change, and Cruise hasn't had time to update the HD map yet.
This can indeed cause system errors, as Cruise's L4 autonomous driving solution uses high-precision maps as an important data dependency.
But whether the scene described by the elder brother is true is not certain.
From an operational point of view, the Robotaxi team involves several links.
The first is the algorithm of the car side, and the complex corner case will indeed cause the system to freeze out, such as the high-precision map information lagged by netizens, resulting in a conflict between sensor data and map data.
If this is the case, then Cruise's L4 solution is far from hot in terms of pre-data fusion.
Even if the car-side algorithm cannot cope, there should be a back-end platform party to take over the vehicle remotely, and the key technical points of this include 5G network, vehicle-side V2X facilities, communication security and so on.
It stands to reason that Robotaxi operators, who have been licensed to go on the road, should have this system to cover.
But the situation was that Cruise eventually sent someone to solve the problem.
So the reasons for the collective downtime of the Cruise team are very likely.
For example, a mismatched fleet scheduling strategy has led to multiple vehicles driving to the same location at the same time.
It is also possible that the entire system was hijacked by hackers.
Or the local 5G network in the United States, Tera, Cruise wants to dispatch remotely but has no heart.
In addition, there is also a possibility that Cruise's "5G cloud driving" system has not been developed at all?
Cruise doesn't say that the outside world will never know.
But there is no possibility, the problem must exist, and it is not a small problem.
Unmanned vehicle downtime, how to deal with it?
Cruise's all-man Robotaxi, high open.
Since January this year, the company has been providing an all-manned operation service in San Francisco, and GM Iron Lady has experienced it herself, constantly praising it with words such as "incredible, smooth, and perfect".
In the course of the operation, Cruise's Robotaxi did catch fire, but not because of its outstanding self-driving capabilities, but because the self-driving system made a jaw-dropping plot.
In April this year, also in San Francisco, a completely deserted Cruise Robotaxi was stopped by the police because it was driving at night without turning on the lights, and as a result, the police did not pay attention and ran straight to the road.
For the first time in history, unmanned vehicles "escaped illegally", leaving the police and legal experts with a complete problem.
And such a situation, together with the large-scale cluster congestion incident of unmanned vehicles, has led to a problem:
How to deal with traffic problems caused by unmanned vehicles?
This so-called "handling" refers to how to reach a set of standard procedures for handling violations between unmanned vehicle development, operators and traffic management departments.
In this process, in addition to ensuring legality and safety, the behavior of unmanned vehicle operators should also be regulated.
This time Cruise's unmanned vehicles inexplicably gathered illegally, causing traffic congestion, and remote dispatch could not solve the problem.
Judging from the results, such a scheme is clearly substandard.
Autonomous driving can never achieve 0 mistakes and 0 accidents, which is already a recognized fact in the industry from both a legal and technical point of view.
Therefore, it is more necessary for autonomous driving companies to have a sound accident response mechanism and technology stack.
First of all, in terms of fleet scheduling, should we design a more reasonable strategy to avoid a large number of vehicles flocking in one direction at the same time?
In addition, the vehicle's own ability to cope with complex scenarios also needs to be improved, which requires AI drivers to learn more atypical scenarios in addition to driving skills on conventional roads.
There is also the security of data communication security in smart cars, which is also a hidden danger that can lead to serious consequences.
So Cruise and Robotaxi still have a long way to go, the first step, may wish to learn from Chinese friends, starting from the "5G cloud driving".