Mobileye believes that human drivers are a very complex group that appears to the computer as "non-standard": there are aggressive drivers and conservative drivers, and it is difficult to accurately predict the intentions of individual drivers. Human beings play games all the time while driving, and unmanned vehicles want to coexist with human pilots and have to participate in this game. The purpose of the RSS theory model put forward by Mobileye is to sum up the conclusions of people's driving knowledge into mathematical models and to teach every unmanned vehicle.
Robots must not harm humans or sit and watch humans be harmed;
Unless the first rule is violated, robots must obey human orders;
Unless the first or second rules are violated, the robot must protect itself.
The birth of all new technologies that break the boundaries is inevitably accompanied by moral ethics and discussions about human survival.
It used to be a robot. It was the turn of the car today.
So far we have done enough discussion on robots. There are also big names in the industry who insist on different opinions on artificial intelligence like Musk and Zuckerberg. At least when strong artificial intelligence arrives, humans will not be totally unprepared. Lamb.
Compared to robots, unmanned vehicles are undoubtedly life-threatening. From the beginning, the constant occurrence of major negative news has been stimulating people's nerves. The unmanned car is like a child who just learned to run and has to accept the trial of everyone.
In March of this year in Tempe, Arizona, faced with a woman who violated the rules and crossed the road, Uber’s unmanned vehicle chose to run over without slowing down; in the same month, Tesla Model X was in assisted autopilot ( In the Autopilot mode, the driver crashed when hitting the barrier; last week, other driver's mistakes caused a Waymo automatic car to get involved in an accident.
The occurrence of these accidents is exactly whose responsibility is very important, but nobody cares about it in the current public opinion environment. Although after nearly two months of investigation, the police thought that the safety officer and pedestrians crossing the road were also responsible for the violation, but the Uber unmanned vehicle, the “first automatic driving car killing†hat, could not be taken off in 1:30. Now. When the Waymo accident occurred a few days ago, many people subconsciously thought it was the fault of the unmanned vehicle before they understood the scene.
The maturity of driverless technology is only a matter of time. However, it is still a thorny problem to make the public believe in unmanned vehicles, share roads with unmanned vehicles, and even hand over life safety to unmanned vehicles. Every accident of a car without a car, even if it is only involved in a car accident, will cause public suspicion and panic. Over time, this distrust can destroy this emerging industry.
Fortunately, someone is already considering this issue and given a convincing answer.
Intel’s Mobileye, a veteran automotive image technology company, published a paper entitled “The Official Model of Safe, Scalable Autonomous Vehicles.†In this paper, the company proposed a theoretical model called RSS (Responsibility-Sensitive Safety).
This theory advocates evading accidents before accidents, rather than waiting for accidents to enter complex culmination procedures.
Intel and Mobileye hope that this theory, like the “three laws of Asimov Robotics†in the autopilot industry, will become a universally accepted unmanned vehicle code that will allow the public to feel at ease.
Mobileye believes that human drivers are a very complex group that appears to the computer as "non-standard": there are aggressive drivers and conservative drivers, and it is difficult to accurately predict the intentions of individual drivers. Human beings play games all the time while driving, and unmanned vehicles want to coexist with human pilots and have to participate in this game.
Taking car parallel roads as an example, a scraping occurs when the road merges. The front car will say that the rear car enters too quickly and the rear car will say that the front car driver is too slow to react. In the end who said it is true, who said it is false , also have to traffic police and insurance companies to decide. Unmanned vehicles have no mouth, and the best way is to act on the premise of ensuring absolute safety.
The true road conditions of Mobileye's aerial photography, the white car that is not getting into the heart
At the Intel Investment Global Summit, Mobilenon CTO and co-founder Amnon Shashua said when he talked about the issue of car accidents:
Under the framework of RSS, if a car accident occurs, it must be the responsibility of the person.
In order to do this, Mobileye also proposed two laws for unmanned vehicles:
First, driverless vehicles never cause accidents on their own initiative;
Second, the driverless car responds correctly to misconduct of other drivers and does not expand the accident.
These two laws of iron should be like the laws of robots that cannot harm human beings and cannot be seen to be hurt by humans. They are suspended above every decision made by every unmanned vehicle.
Sha Shuya pointed out that the mathematical model of RSS can solve three problems: 1. What is a dangerous situation; 2. What is the correct response? 3. Who is responsible for the accident?
First, the sensor fusion of camera, radar, and lidar enables the vehicle to detect the surrounding vehicles, pedestrians, and driving paths simultaneously, and calculates the distance at the same time, so it can better determine the road conditions.
The safe distance calculated in the RSS mathematical model can only be satisfied if the vehicle meets the safety distance requirements.
Then, when taking action, RSS will ensure that it will change lanes, merge lines, and turn when it has road rights, meets safety distance requirements in front of and on both sides, and has no visual dead ends.
Four principles of RSS
Using data-driven machine learning algorithms, Mobileye has conducted 100,000 drive test simulations under the RSS model. The system performance has achieved a frequency response of 10hz per second without any accidents.
When referring to the necessity of the RSS model, Sha Shuya said, “The maturation of technology is only the first step. The thing that RSS does is to sum up the general knowledge of people’s driving knowledge into a mathematical model and to teach every unmanned vehicle. ."
However, if the car is too sensitive, it will lead to another extreme of inefficiency. Passengers are like a well-protected but slow-moving egg that undergoes constant acceleration and deceleration. The same experience is poor. How to solve this problem?
Sha Shuya said, "In the unmanned vehicle industry, there is no such thing as this. In order to ensure the safety of users and other vehicles, what the unmanned vehicle does must do.
There are concepts of false negatives and false positives in statistics. The topics discussed today are 'the brakes are not braked' and 'should not brake but brake'.
The false-negative test is very good, because when the brakes are not braked, the car must hit, only need to set roadblocks, statistics on the passing rate of the unmanned vehicles on it. It is relatively difficult to detect false positives. Can you imagine your unmanned vehicle shaking a bottle of water on the road? No one is willing to take such a car. After training in the RSS system, Intel has collected 200 million kilometers of actual driving data, reducing the probability of false positives to 2.5 per thousand, or only once per 50,000 kilometers.
“As soon as the RSS is accepted by the unmanned vehicle industry, the accident will be more easily condemned. Without this model, there is no way for the unmanned vehicle to really hit the road. Looking at the accident video of Waymo, the crashed car has no place to dodge, but Waymo still took the blame, so we urgently need a model to divide responsibilities, define dangerous behaviors, be accepted by the regulatory authorities, and clearly explain the cause of the accident to the public instead of presenting everyone with a plausible black box."
If artificial intelligence is compared to steam and a human-free vehicle is compared to a train, there is no doubt that what is happening today is a revolutionary traffic revolution like the first time steam trains have started.
Sha Shuya said, “The process of replacing the other cars by unmanned vehicles is a long and gradual process. In this process, priority must be given to the coexistence of unmanned vehicles and human drivers. When the number of unmanned vehicles reaches a certain number, we may be able to Communication between the unmanned vehicles, but that is another story."
In another story, the traffic jam did not exist, and the traffic lights disappeared. The insurance company was unemployed. It was enough for people to sit in the back seat and watch a drone.
08*16 2835 Led Neon Flex,Neon Flexible,Neon Flex Light,Neon Flex Led Strip
NINGBO SENTU ART AND CRAFT CO.,LTD. , https://www.lightworld-sentu.com