What is holding back the implementation of autonomous driving?

　　Self-driving cars are no longer limited to test tracks or quiet suburban streets. They are now seen in places like New York, San Francisco and Pittsburgh in the U.S., participating in real-world traffic, and are also being deployed in countries like Europe, South Korea, Singapore and Japan, honing their skills surrounded by humans and preparing to use their robotic systems to improve chaotic, disorderly street traffic. Let's take a look at the relevant content with the automotive electronics editor.

　　在类似波士顿这样的城市学会如何开车是非常具有挑战性的，那里有颇具创意的左转弯以及貌似随意的让行规则。不过，与发展中国家中司机的侵略性驾驶方式以及城市错综复杂的道路相比，这些都不足为惧了。跟那些对交通信号、警示和限速标志没有一点尊重意识的司机比起来，Patriots 橄榄球队(球队以横冲直撞的打法著称)的球迷看起来都算是正常人了。

　　There are many roads around the world with no lanes and huge, chaotic intersections where pedestrians dominate the traffic and each driver needs to adjust his or her driving strategy on the spot based on the behavior of others, without considering the constraints of traffic rules on himself or herself.

　　These informal transportation systems exist in many areas and cause serious losses. According to the latest data from the World Health Organization in 2013, 50 countries in the world have such extremely dangerous roads, 44 of which are in Africa or the Middle East. In 2013, the number of traffic accident deaths in these countries was close to 250,000, accounting for one-fifth of the total number of traffic accident deaths in the world that year.

　　However, the reality is that the areas that could benefit from autonomous vehicles are unlikely to get the technology any time soon.

　　“A lot of the work we’re doing in autonomous driving right now might not work in third-world countries,” said Ram Vasudevan, director of the Ford Center for Autonomous Vehicles at the University of Michigan.

　　Autonomous driving technology needs to understand the intentions and trajectories of everyone and everything on the road, including vehicles, bicycles, pedestrians, construction workers, playing children, pets, and even darts accidentally fired from toy guns. In the driving environment, people follow a set of established rules, while the expected behavior of autonomous vehicles in the environment needs to be constrained by the law.

　　The fewer traffic rules there are, the more important it is to be able to predict intentions. Because people may make unexpected moves, cars cannot simply rely on common rules to govern their behavior. For example, if other people on the road follow the lane rules, it will only be useful for the driver to park the vehicle in the lane. Otherwise, it will still be dangerous.

　　Driving in the Middle East and Africa is more casual than in suburbs and cities in the U.S. In Lebanon, it’s common to see vehicles driving in the wrong lane, running red lights, or even zigzagging on wide roads out of mischief, ignoring traffic laws.

　　“There are no rules, and anything is possible,” said Daniel Asmar, a computer vision expert and engineering professor at the American University of Beirut. “People can handle this pretty well, even if they get frustrated or honk at each other.” For computers, this chaos is a huge challenge.

　　Vasudevan believes that even in a relatively orderly environment, sudden chaos can cause traffic jams and even accidents, such as when an autonomous car hesitates too long when merging on a highway. This may be due to the car's software, which is reluctant to merge in front of a speeding car for safety reasons, or it needs more time to grasp the surrounding scene and the intentions of other vehicles. If the car is placed on a road where there are no signs or traffic signals and rules for giving way, it usually takes a certain amount of reaction time to improve the system's performance to cope with the current situation.

　　More importantly, self-driving cars also need map data, and many parts of the world do not yet have such information. Self-driving technology requires detailed maps that cover all street information, including height restrictions through streets, detours set up due to temporary construction, and the precise location of street signs and traffic signals in three-dimensional space. Self-driving fleets are already driving in some cities, constantly capturing data, and updating it to develop such maps.

　　In places like Lebanon, where both Google Maps and Apple Maps have some basic flaws, missing data is a huge disadvantage. Even if the world had an incredibly detailed map, it would require extremely intensive maintenance. “In a structured environment, you don’t have to do maintenance very often because not a lot of things change,” Asmar said. “But in an unstructured environment, where everything changes, you can imagine how many times you need to rebuild the platform. It’s a very daunting task.”

　　A handful of wealthy countries in the Middle East have already begun moving toward self-driving. In Israel, where companies are developing crucial self-driving software, the country last month opened its first road for testing driverless cars. In Dubai, a low-speed, 10-passenger shuttle began running along the city’s riverfront business district last year. City officials say the goal is to have a quarter of local traffic without a driver by 2030. Dubai’s police also plan to launch a small, driverless autonomous patrol car by the end of this year.

　　But India and China are two countries that have both driving chaos and local companies working on developing self-driving vehicles. Undoubtedly, their efforts need to overcome more obstacles. According to Bloomberg, Indian company Tata built a test road outside Bangalore to simulate local road conditions, where self-driving vehicles need to compete for the right of way with fearless pedestrians and lost cows.

　　The company still has a long way to go: a senior vice president of Tata told Bloomberg that Tata's computer vision system currently cannot effectively identify 15% of the "vehicles" on Indian roads due to the strange shapes and sizes. (Last year, when former Uber CEO Travis Kalanick visited India, he joked that India would be "the last country on earth to achieve self-driving" because "Have you seen how the locals drive?")

　　Baidu in China is also publicly developing an autonomous driving program. Baidu is working with more than 50 companies around the world to develop autonomous driving software systems. In a previous demonstration video, Baidu CEO Robin Li sat in an autonomous driving car, driving on the streets of Beijing, and made some unsafe operations along the way.

　　Since self-driving cars are not legal on Chinese roads at this stage, Chinese police said they would investigate whether Mr. Li’s actions violated any laws and regulations. (India is also considering a similar ban on self-driving cars due to job losses.) Despite the regulatory hurdles, Baidu President Zhang Yaqin still expressed his confidence to Bloomberg, saying that the company’s driverless cars “will be on the road as early as next year.”

　　Didi Chuxing, a Chinese ride-hailing company, has taken a more cautious approach. Although the company opened an office in California this year to develop driverless technology, its president, Jean Liu, unexpectedly said in a recent interview with Charlie Rose that a "disruptive" switch to autonomous driving would be dangerous. "I think people should pay more attention to how safe this technology is rather than how quickly it can be implemented," Liu said.

　　In China, self-driving cars must learn to deal not only with other cars, but also with electric cars and pedestrians who ignore traffic rules. A Didi spokesperson said that self-driving cars need to understand the difference between traffic signs and traffic signals in different areas, especially in China, where these signs are not as standardized as in the United States or Europe. Therefore, Didi's scale brings great advantages to the company. The company said that human drivers provide 25 million trips a day, generating more than 70 TB of data that can be used for the development of self-driving technology.

　　Driving environments vary in different regions, but autonomous driving companies provide the same data and software

　　Currently, most companies test driverless vehicles by setting up many unexpected scenarios for the cars on controlled roads. For example, Waymo trains its cars in its own mysterious base, Castle, where human testers maliciously block high-speed cars, merge from lanes in blind spots, throw basketballs at vehicles, and so on, using various testing methods to improve the vehicle's response.

　　However, because AI training is based on many assumptions, it often fails when applied to a different scenario. Studies have found that facial recognition algorithms trained on Caucasian test subjects perform poorly when applied to African American faces, and algorithms trained on East Asian test subjects also perform poorly when applied to Caucasians. This difference may also exist in the field of self-driving cars, so related software is always tested carefully and takes into account a variety of extreme situations.

　　But while there are some regional differences in how people drive, it’s unlikely that manufacturers will configure driver software specifically for each region. “Right now, the same data formulas and software are being used across cultures,” said Matthew Johnson-Roberson, a professor at the University of Michigan and co-director of the Ford Center.

　　Most importantly, the car in training will respond to all the information collected from the outside world. An Uber spokesperson said that in order to improve the adaptability of the software, Uber is testing in multiple cities, collecting test information under different conditions and at different times, and the mileage has exceeded 1 million miles.

　　But even if self-driving software can understand unruly drivers and predict how they might break the law, self-driving cars are likely to be limited. A spokesman said Uber's cars will always follow local traffic regulations.

　　Stephan Hoenle, senior vice president of automated driving at Bosch, agrees: “Automated driving can be deployed much more smoothly if it does not violate local traffic regulations.” The driving style of an autonomous car may vary depending on needs and expectations, but breaking the law is not an option - it is a huge responsibility for manufacturers.

　　The problem is that in some places, following the letter of the law can be more dangerous than imitating human drivers who break the law. When impatient commuters change lanes during rush hour, the cars behind them fail to adjust in time, resulting in a series of rear-end collisions.

　　North America, Europe and Singapore will lag far behind developing countries that are in urgent need of autonomous driving

　　For those who work every day to improve the performance of autonomous driving, there seems to be no rush to explore the ultimate capabilities of self-driving cars. "So far, fully autonomous cars have not appeared, right?" said Johnson-Roberson of the University of Michigan. "From an engineering perspective, I don't know anyone who is working on this problem because some of the basic principles have not been realized yet."

　　Delaying the exploration of these issues could slow progress in the regions that need self-driving technology the most. Hoenle claimed that self-driving cars will be widely used in all parts of the world, but he also admitted that it will take some time to develop. He also said: "Compared with the United States and Europe, other regions are generally slower in technology growth."

　　But Carlo Ratti, director of MIT's Senseable City Lab, predicts that developing countries will eventually catch up. "Each new technological frontier is likely to increase existing social gaps," he wrote in an email. "However, subsequent technological diffusion may cause interesting 'overtaking' effects that reduce these gaps."

　　For example, mobile phones were originally available only to wealthy Westerners, but are now common in Africa. Today, African startups are offering new ideas for mobile banking and health services, and Ratti said, “There’s no reason to think that self-driving cars will follow a different path.”

　　The transition zone between the emergence of self-driving cars and the "overtaking on the curve" stage may be a bit long, because they must adapt to their surroundings and need to collect specific data for each street. If they are not designed properly, their development may be curbed.

　　Developers sidestep questions about regional differences, leaving the question to the “growth curve” outside the market. As the core technology for autonomous driving develops rapidly in places like North America, Europe and Singapore, it could lag far behind the developing countries that need it most.

    The above is an introduction to the topic of "What is holding back the implementation of autonomous driving in automotive electronics?" If you want to learn more about this topic, please pay more attention to eeworld. eeworld Electronic Engineering will provide you with more complete, detailed and updated information.