news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Summary:Technology is no longer a wild imagination. If it is to have a smooth landing, it is far from the responsibility of one industry or a few people.

In April this year, at a technology forum in Beijing, lawyer Xing Keke saw legal experts engaged in a debate.

"Traffic laws stipulate that a vehicle can only be defined as one that is driven by a human. Without a human driver, it is at most a driving device!" "94% of the problems with traditional vehicles are caused by humans. Now that machines are replacing humans, who will be responsible?"

At this moment, Xing Keke felt at a loss as the existing laws could not keep up with the development of driverless cars.

As early as two years ago, Baidu launched the first batch of driverless online car-hailing vehicles "LuoBo KuaiPao" in Wuhan. In July this year, when more and more "LuoBo" drove into the streets of downtown, it attracted public attention.

After the novelty wore off, not only legal practitioners, but also technicians, enterprises, and professors were confused. In their vision, driverless cars should have brought about less traffic jams and fewer car accidents. But in reality, people lacked awareness of rules, and the traffic environment was complex and changeable, causing these vehicles to repeatedly stay in the middle of the road, creating new congestion. There was still a gap in determining the responsibility for the safety accidents that followed.

“The social system is not ready yet,” said an expert in the field of transportation. Technology is no longer a wild imagination, and a smooth landing is far from the work of one industry or a few people.

In mid-July, a car that runs on a street in downtown Wuhan. Photo by Feng Rui

The parked car

"Luo Bo crashed his car and hit someone else! I was scared!" Liu Changhua shouted at the other end of the phone.

In April this year, he took a car from Wuhan Economic Development Zone to Dongfeng Motor Corporation. When he was about to reach his destination, he encountered an intersection with vehicles constantly shuttling back and forth. Liu Changhua saw the traffic light turn red, and the car braked suddenly and hit the rear of the vehicle in front. Liu Changhua panicked for a few seconds and found an emergency help button on the screen in the back seat of the car. He quickly pressed it and called the backstage staff.

After the call, Liu Changhua, who was not injured, calmed down and planned to get off the car and wait. He remembered that about five minutes later, a staff member wearing a fluorescent green vest rushed to the scene. Liu Changhua saw that more and more people gathered around to watch the excitement, and the sound of "click, click" of taking pictures could be heard. The staff was busy waving their hands to stop them, and took out a silver-gray cover to cover the "radish" tightly. "This is to cover up the shame!" someone shouted, "We need to cover up the radish's ugliness!"

According to Liu Changhua's description, the owner of the car that was rear-ended saw that the damage was not serious, and finally did not notify the traffic police, but negotiated with the staff to compensate for the repair costs.

After that, Liu Changhua would always keep an eye out for the "radish" around him every time he went out. In the past three months, he had seen the "radish" parked in the middle of the road many times.

He had seen two "Carrots" collide with each other while turning, neither of them giving way to the other. The traffic police on the side felt a little embarrassed and had to evacuate the surrounding cars, leaving the two of them honking their horns at each other; at other times, the "Carrots" rear-ended other vehicles, or were rear-ended by other vehicles.

The warning lights of the "radish" light up from time to time while it is driving. Photo by Feng Rui

At 8 p.m. on July 9, when Chen Lang was driving past the Third Ring Road in Wuhan, the "Carrot" driving in front of him suddenly stopped. By the time Chen Lang reacted, the two cars were getting closer and closer. He hurriedly stepped on the brakes, but they still collided.

Chen Lang got out of the car to check. He was puzzled. The road was so wide and there was nothing in front of him. Why did the "Carrot" stop? He just took out his mobile phone to call 110, and the "Carrot" started forward again, "just running away." Chen Lang smiled bitterly. He admitted that the collision was his fault, but he couldn't help feeling distressed when he saw the dent on the front of his car.

By this time, the controversy had already spread across the Internet.

On social media platforms, thousands of short videos captured the predicament of "Carrot". "Carrot's fast running caused traffic to be paralyzed!" "Silly Carrot got into trouble again!" More than 10,000 comments scrolled below.

The criticism gradually turned to driverless vehicles. On the Wuhan city message board, "sudden braking" and "unexplained stopping" became labels for driverless vehicles. "It seriously affects traffic efficiency." Some citizens complained, "We demand that they stop!"

But in reality, the vision of technology is not like this.

In 2022, Baidu Chairman Robin Li said in a public speech that autonomous driving is safer than human driving, and its perception is much stronger than that of humans, and it can see blind spots that humans cannot see. In terms of operation and execution, machines are also more stable than humans: people get tired, emotional, and drive dangerously, but machines don't.

At that time, according to Li Yanhong's description, if smart cars and shared driverless taxis are widely used, the number of cars on city streets can be reduced by 60%, while reducing road traffic accidents by 90%.

Now that imagination has been realized in reality, why does a contradiction arise?

The screen of "LuoBo" shows the real-time road environment. Photo by Feng Rui

Who is making the mistake?

"The system rarely makes mistakes on its own initiative!" In late July, Li Chen, who works at an autonomous driving technology company, saw the controversy surrounding driverless cars. Faced with the accusation of "disrupting traffic", he felt a little aggrieved.

Li Chen said that the current technology route commonly adopted by unmanned driving is "single-vehicle intelligence". Vehicles will autonomously perceive the surrounding driving environment through cameras, laser radars and other equipment. Once these devices identify obstacles, signals will be sent through pre-designed algorithms to allow the vehicle to make turns, brakes, change lanes and other operations.

"For the system, complying with traffic regulations is a bottom line." Li Chen emphasized that in the early stages of system design, technical personnel will write every traffic regulation into the code and review it before the vehicle is put into use.

In order to obtain an operating license, every driverless car needs to pass a series of "exams".

Li Chen's company has cooperated with many local governments. According to the regulations of a local government, an unmanned vehicle must first be tested in a closed area designated by the government, and equipped with a safety officer sitting in the main driver's seat. Li Chen explained that the presence of a safety officer in a vehicle does not mean that the system has not yet reached the "unmanned driving" standard, but is due to the requirements of government supervision. In the event of an emergency, the safety officer can take over the steering wheel and operate.

After that, each driverless car must run 5,000 kilometers without any "active mistakes" leading to accidents before it can be promoted to the next stage: let the safety officer sit in the passenger seat and only step on the brakes when necessary. After passing the test again, the safety officer in the car can move to the backstage to remotely monitor the vehicle. At first, one person watched one car, and gradually one person watched five or ten cars.

"It's not like a driving school simulation test, where you hold the steering wheel all the time, but you are responsible for issuing instructions in special circumstances, and the execution is still left to the system itself." Li Chen explained that the task of the backend is to deal with problems that technology cannot solve. People are not willing to "overtake", and the government will count the number of times that humans take over to assess the technical level of the company.

He saw that it took several months for a self-driving car to pass the test in some cities and 300 days in others.

In Wuhan, the government has designated several testing areas. Photo by Feng Rui

However, it is obviously not enough for the system to just follow the rules.

Li Chen observed that, contrary to the system's "bottom line awareness", people often regard traffic rules as "moral behavior" that can be occasionally violated.

When Li Chen was riding in an unmanned vehicle, he saw a pedestrian crossing the zebra crossing and suddenly standing in the middle of the road. Fortunately, the system stopped in time to "give way to the pedestrian". The pedestrian was still looking down at his phone until the traffic light turned red. The system had to scan repeatedly through the camera to determine whether the person had moved.

"The machine's reaction speed is 100 milliseconds, while a normal person's reaction speed is about 300 to 400 milliseconds." But Li Chen admitted that sometimes even the system cannot avoid mistakes made by pedestrians and other cars in time.

At Li Chen's company, driverless vehicles still encounter some "passive mistakes" accidents. Once, while the driverless car was driving, a driver on the opposite side quickly crossed the lane and tried to merge. This time the system did not react in time, causing the two cars to collide with each other and the front of the car was scratched.

Li Chen smiled bitterly. At this time, the company could not help feeling anxious. One accident could very likely make ten thousand safe trips come to naught.

In Wuhan's "Carrot Run" Automotive Robot Intelligent Valley, a large number of "carrots" are parked waiting to be operated and repaired. Photo by Zhou Yufan

Difficult to adapt

In the view of Dai Jiarun, a teacher at the School of Computer Science and Technology of Fudan University, a system that is "afraid of making mistakes" will find it more difficult to adapt to complex road environments.

Dai Jiarun and his team once conducted simulation tests on the technology of driverless companies: by simulating city scenes on computers, they observed the difficulties that driverless cars might encounter during driving.

They found that the system's choices tended to be more "conservative" than those of human drivers.

During the test, there was a flattened cardboard box in the scene. In the eyes of the testers, "It's just a box, so why not just step on the accelerator and run it over?" At this time, the driverless vehicles could not determine what the cardboard box was, or whether there were other dangerous objects inside. As long as there was a possibility of collision, they would generally stop where they were.

Some driverless car companies have explained to Dai Jiarun that when designing algorithms, they would rather slow down the vehicle and stop when encountering obstacles. They believe that reducing efficiency will make the vehicle safer and the implementation of technology more stable.

A self-driving model car equipped with laser radar and cameras at the School of Computer Science and Technology of Fudan University. Photo by Zhou Yufan

Nevertheless, the diversity of road scenes still exceeds the imagination of designers.

Li Chen has seen a driverless vehicle encounter a temporary road closure when passing a small road. The system's judgment of these construction sites is usually based on information from high-precision maps, but this situation is not shown on the map at this moment. Faced with a whole row of roadblocks, the driverless car was a little "at a loss".

At this point, the system had no choice but to ask for help from the remote safety officer. Li Chen remembered that after the safety officer discovered the problem with the system, he gave the command to turn around. The driverless car then began to determine if there were other pedestrians or cars around, then turned around and chose a new route.

The contradiction between the system and the roads is more prominent in the city center.

"For a long time, driverless cars will have to coexist with ordinary vehicles." Dai Jiarun said that people's driving habits vary greatly. During the morning and evening rush hours, some drivers move slowly, while others keep cutting in line. As the traffic becomes larger and larger, driverless cars will hesitate to change lanes or turn to prevent collisions, and even drive at low speeds in the fast lane and stop directly to deal with other drivers' "cutting in". Some decisions that ensure safety have instead created traffic jams.

"Autonomous driving is not as smart and flexible as humans." Dai Jiarun said that in the face of complex situations, it is difficult for algorithms to strike a balance between "conservativeness" and "efficiency." He is worried that in the next one or two years, when autonomous vehicles enter more cities and operate at all times, the probability of congestion and accidents will increase further.

But Li Chen saw that the system was growing despite setbacks.

He said that the government itself has some restrictions on the proportion of driverless vehicles. A local government once stipulated that driverless licenses cannot exceed 10% of all operating vehicles, and can only be increased to 20% after the system is gradually optimized.

Li Chen said that driverless vehicles will constantly collect extreme situations and input them into the system for training, so that the vehicle can form "muscle memory". Although the probability of these extreme scenarios occurring is only 0.1%, companies still have to invest more than 80% of the cost for this.

"The difficulties on the road are endless." He sighed. Even if the system has reached the most perfect version and there are staff in the background to take over, it cannot cover 100% of the scenarios.

In Yang Min's team at Fudan University, where Dai Jiarun works, autonomous driving simulation tests will identify extreme scenarios where accidents may occur and classify and analyze them. Photo by Zhou Yufan

Why to judge

Once a safety accident occurs, how should responsibility be determined?

As early as 2020, an automobile company approached Xing Keke to ask, if the company wanted to make driverless products, "what if a traffic accident occurred?"

Only then did Xing Keke realize that although the technology for driverless cars was already so mature, there was a gap in the relevant legislation.

Xing Keke pointed out that according to the current Road Traffic Safety Law, which is my country's basic law in the field of road traffic, the main top-level design is centered around "people" - the driver of the vehicle. There is no separate provision for driverless driving, and the law does not clearly define who is responsible if a traffic safety accident occurs.

"At present, unmanned driving is still a form of autonomous driving." Xing Keke said that according to the current common way of defining responsibilities, the judgment of responsibility will be different depending on whether the system made a completely autonomous decision or whether a person intervened in the vehicle to varying degrees when an accident occurred. "If a driver or remote safety officer makes a mistake when operating the vehicle, it is not considered completely unmanned driving, and the operator may need to bear responsibility for the accident. If the accident occurred in a completely unmanned state, the responsible party may involve the vehicle owner, manager or system manufacturer."

Xing Keke has seen that in recent years, cities such as Shenzhen and Shanghai have passed local regulations to try to clarify the responsibility for driverless accidents.

Despite this, the "connection" between regulations has become another major problem.

Xing Keke remembers that everyone mentioned in the forum that local regulations should not conflict with the more powerful Road Traffic Safety Law, Civil Code, Criminal Law, etc. Some local regulations state that "when an unmanned vehicle causes a traffic accident, the public security organs must handle it according to law." At this time, the public security organs' law enforcement basis is still the Road Traffic Safety Law. "If the superior laws have not been changed, how can they guide local regulations?"

"The regulations between different regions must also be connected." Xing Keke pointed out that if different regions have different standards for the access and accident identification of driverless vehicles, it will be difficult for driverless vehicles to be tested and operated across regions in the future.

Xing Keke sighed, "The law is supposed to be a guide for technology, but now it is waiting for technology to grow wildly before making changes, which creates obstacles to the development of technology."

While legislation is still in transition, there are also workarounds in reality.

Xing Keke conducted research on this. When a police officer arrived at the scene of the accident and saw that there was no driver in the car, he was confused for a moment. "We can't just catch a car and fine it, right?" Later, the traffic control department came up with a solution. They contacted the company to call the staff who supervised the driving of the vehicle in the background, that is, the cloud receiver, to the scene and let him be responsible for the subsequent handling of the accident.

Li Chen said that when an accident occurs in the company, the fleet will come forward to solve such problems. At present, the traffic police will still follow the method of handling ordinary vehicles, retrieve the surveillance, check who is mainly responsible, and then impose penalties. "This is a highly regulated industry." Li Chen said. According to regulations, if an accident occurs during the test of unmanned driving, it must be reported to the government department. The company had previously thought about concealing it, fearing that it would leave a "criminal record" for the vehicle, but the vehicle was equipped with 20 or 30 sensing devices, which recorded everything clearly, and there was no way to "escape" at all.

Outside Wuhan's "Carrot Run" car robot Zhixing Valley, staff members are guarding a whole row of "carrots". Photo by Zhou Yufan

Systematic things

"It's not enough to just have a smart car," said Yang Xiaoguang, director of the Intelligent Transportation Systems Research Center at Tongji University. For technology to truly integrate into the city, it is a "systemic thing."

In the 1990s, Yang Xiaoguang first learned about "intelligent transportation system" when he was studying in Japan.

In New Jersey, the United States, there is a self-regulating traffic signal control system. The 128 traffic lights communicate with each other and automatically adjust the duration of the traffic lights according to the current traffic conditions, which can clear up to 8 kilometers of congestion within 15 minutes.

Japan has installed 1,600 roadside devices on highways across the country, which interact with vehicles in real time to inform them of current safety hazards. Academic organizations, industry associations and the government work together to participate in the development of the system.

The city of Hamburg, Germany, has installed sensors on traffic lights that can detect pedestrians waiting at intersections and transmit coordinate information to buses that are about to turn right, preventing collisions between pedestrians and vehicles due to visual blind spots.

Some cities in China have installed traffic lights with sensing devices. Image source: Xinhua News Agency

Yang Xiaoguang said that over the past decade, China's driverless technology has developed rapidly, and companies and industrial and information technology departments want to quickly promote driverless products so that the industry can become popular.

The construction of the smart transportation system is still on the way. Yang Xiaoguang is somewhat helpless. Enterprises are not so familiar with areas other than technology. Most of the "driverless" they talk about is just a type of transportation tool. The entire transportation system is not ready for the implementation of the tool: people's traffic behavior is not standardized enough, and the road traffic environment has not yet reached the standard and specification. More importantly, the corresponding laws and regulations and management are waiting to be improved.

But for this systematic project, relying on the efforts of a few is far from enough.

Li Chen said that the company needs to deal with many government departments: to modify the driverless car, it needs to contact the Ministry of Industry and Information Technology; the camera in the car involves user privacy, so it needs to contact the Cyberspace Administration of China; which roads can be deployed with vehicles, and how to judge the responsibility in the event of an accident, it needs to communicate with the Transportation Commission. The traffic police, housing and construction departments, etc. are also involved in road management and maintenance. "This is a brand new thing, and no one has experience." Li Chen said that some local governments have set up special offices for this purpose, while other areas can only be led by the municipal government, and specific matters are still coordinated by various districts and departments.

"Objectively, there is no competent department for transportation." Yang Xiaoguang pointed out that the functional divisions of various administrative departments in the construction and management of transportation are inherently complicated. Some functional personnel told him frankly that the transformation of transportation involves the division of responsibilities and high costs, and it is better to do less than more. "The coordination of various departments and industries is currently lacking."

Yang Xiaoguang still remembers that when he attended the Shanghai World Expo in 2010, everyone was already imagining that "the cars of the future will be as free as schools of fish in the deep sea, and can safely pass through intersections without traffic lights. Cars can even drive automatically and choose the best route, allowing the elderly and disabled to travel alone."

He smiled and said that some wishes have already been realized, and more imaginations are not far away. Technological development is not only about running faster, but also about walking steadily, which will eventually promote social changes and bring people a sense of security.

(Liu Changhua, Chen Lang, and Li Chen are pseudonyms in this article)