“This carbon was picked up when my team and I investigated the mine.” Yang Wenli, CEO of Lingjun technology, said, pointing to a stone in the conference room. In order to determine the landing direction of autopilot, he and the senior management team went to mines, ports and sanitation companies, and even went to Mohe to investigate the application of tractors on large farms. As the CEO of an autopilot company with only 50 people and the former chief architect of Baidu deep learning research institute, Yang Wenli bluntly said, “large companies are more suitable for market cultivation”. While waiting for the market cultivation of large factories, Lingjun technology has been groping for six years by “slow heat”. In the past six years, the automatic driving industry has always been above the tuyere. The story of Lingjun technology is only one of the optional paths for start-ups, but the story collection supports the trillion market in the future.

2 out of 10 exclusion methods

Yang Wenli, an automation major from Tsinghua University, is no stranger to the industrial scene. When he was a doctoral student in the United States, he went to a mine in Arizona for internship and was very familiar with the internal rules and vehicle driving specifications of the mine. Back home, because of an automatic driving project, he went to a mine in Tongliao, where he picked up a piece of carbon and put it on the table in the conference room all year round.

Tractor factory, farm, environmental sanitation company, automobile production line, port… These scenes have also been investigated by Yang Wenli and his team. In addition, after consulting people in the industry, Lingjun technology has studied all landing directions of automatic driving, a total of 10, and finally determined in the second half of 2021: robobus and urban logistics.

“We visited a car production line in Yizhuang. It is a customized production. Each car has different models, configurations and colors. A robot will feed with the car, and the degree of automation is quite high. There are floor sweeping robots and storage robots, which can build maps, navigate, plan and avoid obstacles independently. Is it automatic driving? I don’t think so. The definition of automatic driving in the industry is related to traffic and should be driven Put it on the road. “ Yang Wenli explained.

For example, this kind of self driving service is also excluded by other companies in Beijing.

“According to the industrial line, smart Walker has three directions: vehicle regulation level, commercial level and special application. Now smart Walker has a higher degree of commercialization in environmental sanitation and belongs to our commercial product line.” As for the reason, the answer of Wang Sunan, vice president of zhiwalker strategy, is similar to Yang Wenli’s logic: “this is determined by the market. The maturity of the industry itself determines whether we will choose the corresponding host factory for cooperation in this direction.”

What’s more, she and Yang Wenli both mentioned “dimensionality reduction attack” and “these three industrial lines use the general software bottom layer of smart walker. We will equip the hardware system that needs to be installed at the bottom layer of the software according to the characteristics of vehicles used in different types of scenes, so as to form the shipping product of smart Walker - driverless brain”.

There are also companies that force the industry to mature, including Baidu, Yang Wenli’s old owner. At the end of 2013, he completed the interview within one month, resigned from the position of chief architect of Western data, and returned home to join Baidu deep learning research institute. Later, baidu cooperated with a number of host factories, established an intelligent driving business group, launched Baidu Apollo, and invested in automatic driving teams in many places.

Yang Wenli is looking for opportunities in small projects of Lingjun technology. In his opinion, “with the current volume and resources of Lingjun technology, it is unrealistic to make the mass production of automatic driving L4 on passenger cars in the short term. We need some commercial landing in the medium and short term”.

Start from operation

Generally speaking, companies that start with the general underlying technology of automatic driving have two commercialization modes: one is to sell cars and schemes for projects, and the other is to put vehicles into operation and be responsible for operation. Most companies such as Baidu, Lingjun technology and Wenyuan Zhixing do both. Specifically, Lingjun technology + Ganzhou, baidu launches robotaxi and robobus in Yizhuang, and Wenyuan Zhixing launches robotaxi and robobus in Guangzhou, which belongs to the latter form.

There are also companies that want to avoid operations and focus on products.

“At present, the company of automatic driving track shows such a current situation in the choice of development path, either controlling the operation end, or controlling the road section, transportation capacity, or controlling the scene. This leads the company to consider the operation end, take down this task, and then complete it through its own products. However, at this stage, because the safety officer cannot be removed, the upgrading and transformation of the original vehicle must be completed through a higher cost, and can not be completed Realize the business closed loop of cost reduction and efficiency increase. Faced with the goal of commercialization, smart Walker chose not to be an operator or vehicle manufacturer with the concept of automatic driving, but to focus on becoming a company with strong product power and cooperate with various industry partners as an automatic driving technology provider. “ Wang Sunan explained.

The market is still immature and there is no right or wrong path. As He Qi, senior analyst of Analysys auto travel, said, “both have their own advantages. Simply providing an autonomous driving brain makes the scheme easier to copy and bears lower operating costs. However, due to the increase in the cost of data acquisition by enterprises not participating in the operation, the training algorithm is constantly supported by operating data”.

On the other hand, most of the cost comes from labor in the mode of vehicle investment + operation. “The technical team of Lingjun technology Ganzhou has more than a dozen people, who are mainly responsible for delivering the R & D products to the site and then doing test operation.” In the work list of Lingjun technology coo Pang Dongjun, there are also investment projects of Suzhou government being contacted, mainly robobus.

Abacus of “golden oil”

Recruitment requires capital and office space. Financing is the first task Pang Dongjun received after he came to lead Jun technology. “I also study automation, which may be one of the majors with the most background of artificial intelligence engineers. If I want to get financing, I have to clarify the technology and implementation with the investor. This is my best field. I can’t get early projects and cooperation by sales alone. If the project in Suzhou is finalized, I need to recruit people”.

This enrollment expansion plan of Lingjun technology is rare before 2021. When the company was founded in 2016, Yang Wenli and two Tsinghua alumni, plus several shared office seats and a whiteboard, became everything of Lingjun technology. The whiteboard was still hung on the wall of Lingjun technology conference room. Now Yang Wenli has found a new office, “the newcomer is coming and needs a larger office space. The decoration costs three or four