This article is originally published by AI Frontier. The original link is t.cn/RTEQOGK


At what point did you think autonomous driving would be a reality?

What Elon Musk wants to do always feels too far away, and there are huge difficulties and obstacles in achieving it.

But starting in 2018, Audi will gradually introduce the technology for autonomous driving in production models, starting with the A8. Baidu has also announced plans to mass-produce driverless cars next year. Autonomous driving technology has moved from the lab to mass production in just a few years.

There are hundreds of self-driving companies around the world, and more than 240 start-ups are involved in the industry to some degree.

The more cutting-edge the technology, the faster it actually advances.

Beijing yesterday became the first city in China to introduce local regulations for autonomous driving tests. As to whether Li should be fined, please refer to the details of the regulations issued this time.

Behind the prosperity of the industry, academia is also closely following. Tsinghua University’s School of Cross-information will offer the course, taught for eight weeks by Lou Tiancheng, who graduated from Tsinghua’s Yao class and founded autonomous driving company Pony.ai, along with a team of experts from his company.


This course covers the overall technical route of autonomous driving and the design and algorithm of major modules, including:


1. Introduction: The overall architecture of the autonomous driving system, the development status and prospects of the autonomous driving industry;


2. Hardware: sensor (multi-sensor fusion), computing hardware, GPS hardware;


3. Perception: Computer vision technology and deep learning;


4. Map: high-definition 3D modeling of road information and static elements;


5. Positioning: positioning based on differential Global Navigation Satellite System and computer vision;


6. Decision planning: global optimal path selection and local motion trajectory planning;


7. Control: use feedback control mechanism to control the vehicle to achieve accurate driving action;


8. System architecture and simulation system: system reliability, stability and real-time, including tests in actual roads and simulation environments.


What is the background of Pony.ai?

On pony. ai’s website, we see the following introduction:

James Peng, PhD from Stanford University, is the former chief architect of Baidu. He is fully responsible for the technical direction of Baidu’s unmanned vehicles, as well as Internet realization technology, big data and infrastructure. Worked at Google headquarters for seven years. CTO: Lou Tiancheng, ACRush, the first programmer known as “Lou Master”. TopCoder ranked first in China for ten consecutive years and won the champion of Google Global Programming Challenge twice. Former Baidu chief architect of unmanned vehicle, Baidu’s youngest T10 engineer. He worked on Google’s self-driving car.

At the same time, Pony.ai gathered a number of TopCoder or ACM contest programming gods, it is said that one of the gods felt that the previous system was not working, and went to rewrite the whole system, can be said to be full of geek style.

They believe driverless cars are getting closer:

As a self-driving company, it certainly needs all kinds of software engineers. Is it impossible for ordinary people to get access to this industry?

As you can see from their job postings above, they are almost all software engineering-related. “You don’t need to have experience in driverless cars, you don’t need to know machine learning (many of the jobs are systematic, not just machine learning), and you don’t need to have competed,” says one Internet blogger who joined Pony.ai after working as an HBase engineer at Xiaomi. “The company basically writes C++.” “of course, unmanned cars are also a big data industry in nature, which inevitably involves data storage and calculation.”


How would you evaluate this course?

Pony. Ai CTO Lou Tiancheng said in a speech that the accuracy and reliability of L4 autonomous driving technology must be at least 99.9 percent, and a delay of 50 milliseconds could lead to fatal accidents. To have a “usable” driverless technology, it is difficult enough to be challenging, and the technology must be constantly improved.

“What about the direction of autonomous driving? I can only say that it is very hot now, all the big companies are investing a lot of human resources and financial resources to do it, I also believe that in the next few years will be able to make very mature products, “” is it safe? You see computer all develop so many years, return total dead machine…… In case it’s on the road, the system goes down…… “. He sees autonomous driving as a great application for video understanding, which is still very much in its infancy and has a long way to go. “In that sense, using the autonomous driving class as an example shows that there are a lot of unsolved video understandings behind Fancy’s technology, Or the theoretical and technical bottlenecks of deep learning — I guess that’s why we started this course.”

Professor Yao Qizhi, founder of Yao Class, has said that China’s current AI research and development is flawed in two ways: first, there is no large AI system, and second, the algorithm and theory behind it are weak. The latter involves the topic of talent. “The combination of production, learning and research is also a way to produce theoretical methodology”.

Under this trend, Tsinghua has the responsibility to meet the needs of the market and cultivate precision talents. As a student of Professor Yao Qizhi, Lou Tiancheng is also a kind of feedback.

In fact, tsinghua’s offering of automatic driving courses partly reflects that we need more professionals. Not only the academic world needs to know about automatic driving, but also the industry. So what do you think about learning to drive autonomously?

Welcome to discuss in the comments!


Automatic driving knowledge popularization

In fact, this course of Tsinghua university is also a basic literacy course. From the outline, it mainly covers hardware, perception, map, positioning, decision planning, control, system architecture and simulation. We are also here, to send you a programmer introduction to unmanned driving free course, one hour of popular science, simple and effective to explain the relevant knowledge of autonomous driving.

The text is the same as the following article:

If driverless Cars are the Future, How do Programmers Get started?

Follow our wechat account “AI Front “and reply to “AI” in the background to obtain the SERIES of “AI Front “PDF e-books