Nearly 30 years before Google, Tesla and Uber started autonomous driving, a team of engineers led by a German scientist named Ernst Dickmanns had successfully achieved autonomous driving.
Today, the director will talk about Dickmanns' autopilot and how he was forgotten. By the way, the ups and downs of autonomous driving in the historical wave of AI.
Dickmanns, 82, no longer gives general advice to researchers. He said, "People should never completely ignore the methods that have been very successful." .
Before becoming "the actual inventor of autonomous driving", Dickmanns spent the first ten years of his work analyzing the trajectory required for the spacecraft to return to Earth in the atmosphere.
As an aerospace engineer, he was promoted rapidly in the ambitious aerospace industry in West Germany. 1975, under 40 years old, working in the research institute of the university.
During this period, his life mission changed: he hoped that the car would find its own way. Dickmanns is increasingly convinced that his future lies not in space, but on land. A few years later, he bought a van, installed a computer, camera and sensor, and began to test it in the university of 1986.
Colleagues at the university said he was a freak, but because of his previous achievements in the aerospace field, he turned a blind eye to him.
1986, Dickmanns' van became the first self-driving car. The following year, the car was tested at a speed of 90 kilometers per hour on an undisclosed highway in Bavaria. Soon after, German automobile manufacturer Daimler found Dickmanns, and Daimler became his sponsor. In the early 1990s, they put forward a project. They want a large bus, which can demonstrate autonomous driving at 1994 Paris exhibition.
After hearing this request, Dickmanns took a deep breath and told Daimler that "my team and our method can do it".
1994 10, Dickerman's team received a group of senior guests from Charles de Gaulle airport. On the highway, they use autonomous driving to complete the driving task. Behringer, an engineer, sits in the driver's seat of the vehicle with his hands on the steering wheel to prevent problems. After 24 years, Berinda still talks about it excitedly.
It was reported in many newspaper headlines. A year later, the team tested another new car. On the highway from Bavaria to Denmark, they drove 1.700 km by self-driving, and the top speed exceeded 1.75 km/h.
However, soon after, the project was forced to stop.
Dickmanns's technology reached the ceiling, Daimler lost interest, and then the pioneering job opportunities in Dickmanns were forgotten.
The history of artificial intelligence has been more than 60 years, and it is often accompanied by a "midsummer" and another "winter". The Dickerman project started in midsummer and ended in winter.
In the late 1950s, the research direction of artificial intelligence was how to make machines think like people. At the beginning, the biggest characteristics in the field of artificial intelligence were thunder and little rain, various propaganda and various platforms. In 1960s, economist herbert simon predicted that "machines will replace human work in 20 years".
These stimulated investment and a large amount of hot money poured in, but technology could not be realized. In the mid-1970s, the bubble burst, funds decreased and artificial intelligence technology precooled.
In the mid-1980s, Dickmanns' self-driving project caught up with a new round of midsummer, and his concept aroused the interest of capitalists. His team once expanded to 20 people. Then came the winter of the 1990s, which made Dickmanns' concept unattractive.
Behringer, an engineer sitting in the cab at that time, said, "This is an interesting concept, but for many people, it is too advanced." .
Technical experts say that there are two kinds of inventions: like light bulbs, which have been used and improved since the invention. Another kind of supersonic passenger plane, such as Concorde, embodies revolutionary technology, but it is too advanced to be suitable for now.
Dickmanns' autopilot belongs to the second category.
When he started research and development in the 1980s, computers still needed 10 minutes to analyze images, respond to the surrounding environment and realize driving.
Facing seemingly insurmountable obstacles, he drew inspiration from human biology. He believes that cars should be as aware of the streets and surrounding environment as people's eyes. Humans can only pay attention to the high-resolution visual center. Similarly, cars should only focus on things related to driving, such as road markings. This will greatly reduce the amount of information processed by the computer.
Auto-driving on the expressway is actually relatively simple, because the road surface is not complicated, the lanes are clear and the signs are clear.
At that time, there were some problems, such as the car in front blocked the sign, or the sign was too old to see clearly, which would lead to problems with automatic driving.
In the cold winter, Daimler told Dickmanns that he hoped to start the research as soon as possible, but Dickmanns knew that the research could not be commercialized for several years. Daimler gradually lost interest in continuing.
JürgenSchmidhuber, co-director of Dalle Molle Institute of Artificial Intelligence in Lugano, Switzerland, said, "In hindsight, it may be a mistake that these projects did not continue immediately. If they continue, there will be no such things as Google and Uber in the field of autonomous driving. "
German companies now hold more than half of the patents on autonomous driving technology, but new participants such as American technology giants such as Waymo of Alphabet are catching up.
In the late 1990s, Dickmanns signed a four-year contract with the US Army Research Laboratory. In addition to a new generation of self-driving cars, their cooperation can handle more complicated roads. When Dickmanns retired, this project attracted the attention of Darpa, the emerging technology department of the Pentagon. So in 2004, the department launched a series of "challenges" to encourage scientists to participate in autonomous driving competitions.
In 2005, Sebastian Trent, a computer professor at Stanford University, won the challenge and established the Google Autopilot Team. He is a celebrity in the AI world. At the same time, Dickmanns, the protagonist of this article, and his pioneering work have also been forgotten.
20 1 1, 17 After the demonstration of the unmanned car in Dickmanns, The New York Times reported that Sebastian Trent had tried to build the first unmanned car, but immediately revised it, saying that Sebastian Trent had developed a driverless car, but he was not the first person to do so.
20 18 with the hype before artificial intelligence, has a new winter come? Many people think so.
Recently, deep learning has been widely used in the research of AI and has been recognized by the algorithm. The basic principle is to find correlation in complex data, which is feasible for most applications, but in some cases it is a dead end. Because deep learning is data-driven, it is accompanied by limitations.
I translated Piekniewski's article "AI Winter is Coming" in computer science in San Diego before. He said that a large amount of hot money poured into AI, especially in the fields of autonomous driving and robotics, and it was unrealistic to expect immediate results.
Many investors are very upset, and they have invested so much money, but their expectations have not come true.
Virginia Dignum, a professor at Delft University, said, "AI personnel will continue to focus on deep learning, and the results will be disappointing. If you want to make a breakthrough in this field, you must build a model through less data or causality, rather than relying on the correlation brought by a large number of data. But she doesn't believe that the winter of AI is coming, because there are a large number of commercial AI, which is due to the technological progress since 20 10, especially the progress of computing power and data storage.
Dickmanns still believes that it will take more than ten years for autonomous driving to really mature. Nowadays, cars really have no resolution. They only rely on a lot of data for training, which means that they will be good on some roads and in general, but accidents will happen in some specific environments.
Dickmanns said, "My autonomous driving technology is named Pathfinder Vision. At present, there are still several institutions studying this method, which can make the car run anywhere and in any situation, whether after a storm, an earthquake or a war zone."
He predicted that one day, the industry will realize the limitations of the current general method, and then he will come out again.
He added, "I'm glad I can be one of the pioneers, but if I start over with the existing technology today, it will be a completely different story."