Didi Zhang Wensong: Technology changes life, using big data to "solve traffic jams"

Didi Zhang Wensong: Technology changes life, using big data to "solve traffic jams"

【51CTO.com original article】

Zhang Wensong: Challenge difficult problems and use technology to do more valuable things

Why did you choose the IT industry? Most people would probably say: because of the money. Yes, it is an indisputable fact that this industry offers good treatment and high salaries.

But for Zhang Wensong, whether it was founding the open source project LVS during college, working at Alibaba for seven years, or rejecting Google and Facebook and joining Didi Chuxing last year, they all adhered to the same original intention: to challenge difficult problems and use technology to do more valuable things.

[[191938]]

Zhang Wensong

Senior Vice President of Didi Chuxing

Zhang Wensong, founder of Linux cluster project - LVS. Currently the senior vice president of Didi Chuxing, responsible for basic platform work and the general manager of smart transportation projects. He worked in Alibaba for nearly seven years, serving as senior technical director of Taobao, chairman of Taobao Technology Committee, vice president of Alibaba, chairman of Alibaba Open Source Committee, CTO of Alibaba Cloud, etc.

The seeds of open source sprouted accidentally

In 1995, Zhang Wensong was still a doctoral student at the National University of Defense Technology. He found that the load balancing equipment used in the construction of Changsha Information Port was extremely expensive. At that time, he had a bold idea - can load balancing be done on Linux?

Just do it! In just two weeks, the code of the open source software LVS was completed. At that time, there was no content related to load balancing in the Linux field, and this project was very practical, so after it was put online, people responded quickly. Not only did they send in requirements, but also functional patches sent by contributors. With the coming and going, the LVS project was iterated quickly, and the functions became more and more complete.

This experience allowed Zhang Wensong to communicate with the world's top technical talents and experience the charm of open source. Therefore, when he later joined Taobao, he always advocated open source and created a low-carbon network through technological innovation.

If the LVS project is a point

Then Taobao is a surface

Joining the more challenging Taobao, Zhang Wensong single-handedly built a core system team to provide numerous basic services to Taobao users and support the entire business.

The new system planning starts from hardware customization, to operating system, and then to stand-alone software optimization, such as Tengine, JVM and database optimization.

The database is the biggest bottleneck of the entire trading platform. How to use Taobao business to IOE to cope with hundreds of thousands of transactions per second, as well as CDN systems, distributed storage, distributed cache systems, big data platforms, image search, etc., but these are only a small part of what Zhang Wensong has experienced at Alibaba.

Taobao is the largest online shopping platform in China. People are involved in the selection and final decision-making of the entire online shopping process. At most, they make recommendations or advertisements during the selection process. How can we use technology to achieve more valuable things?

Artificial intelligence is here, and so are new challenges

At this time last year, Zhang Wensong joined Didi, and Didi’s transformation in the transportation field began.

When we use Didi, we just need to input a destination and we can sit back and relax. However, behind this seemingly simple UI, everything relies on calculations. How to achieve perfect and effective matching is a very complex calculation problem, and it also requires extremely high real-time performance. All of these use artificial intelligence algorithms.

"Didi Traffic Information Platform" Beijing Road Traffic Heat Map

Traffic forecast

Based on a large amount of real-time reported trajectory data, the real-time road conditions can be calculated, and future road conditions can be predicted based on historical road condition data and real-time road conditions.

Path Planning

Plan the routes for the starting point and destination on the entire road network, including the pick-up route and the drop-off route.

ETA calculation

Use ETA technology that incorporates extensive machine learning to estimate the time for both pick-up and drop-off.

Intelligent order splitting

The most ideal dispatch matching is not to send the nearest driver to accept the order as soon as the passenger places an order. Doing so may result in another passenger's car demand being more consistent with this driver 1 millisecond later. In view of this situation, the passenger's order demand is first placed in the demand pool. Didi sets the matching time to 2 seconds per time to take into account the user experience.

Intelligent order distribution constructs a huge supply and demand matrix, with the passenger order demand list vertically and a large number of available drivers horizontally. The platform needs to calculate the correlation between users and drivers.

There are many dimensions that can be used as references for correlation, such as physical location, location mapping to road network conditions, pick-up time, driver service score, passenger preferences, etc. These multi-dimensional items are weighted and mapped to a space between 0 and 1. The larger the value, the higher the correlation between the passenger and the driver. This process must be controlled within 100 milliseconds. Because the unloaded driver is constantly driving and his location changes at any time, real-time performance is a factor that must be improved.

In addition, there is also the problem of imbalanced supply and demand in different areas of the city. For example, if there are passengers A and B, and the driver is closer to A and slightly farther from B, the driver is likely to be assigned to A based on the 2-second matchmaking. However, if destination A is in oversupply and destination B is in short supply, from the perspective of global supply and demand balance, it may be more reasonable to assign the driver to B first.

Using technology to do more valuable things - smart transportation

As for the entire online car-hailing platform, due to the business characteristics of route settlement price, it is necessary to report GPS positioning point data to the platform in real time. These positioning points are connected to form the driving route. Millions of vehicles travel on N routes every day, which can well reflect the urban traffic conditions in China. This is a new direction that can bring social value.

  • In the past, most traffic systems used detectors, coils, checkpoints, cameras, etc. to collect data at various intersections, which had a long cycle and poor timeliness.

  • Now, Didi’s massive data is used for traffic management, operation, supervision, and decision-making, and the best routes are recommended to consumers, bringing more convenience to users’ travel .

Didi will focus on smart traffic management, smart operations, overall traffic measurement system construction, and smart travel for all travelers. Didi will use its advantages in IoT, big data, cloud computing (Internet of Vehicles platform - mobile Internet - big data platform - artificial intelligence algorithm - cloud computing platform) to "solve traffic jams" and bring changes to the entire transportation field.

For example, the Didi smart traffic light project that is currently being implemented adjusts and optimizes traffic lights based on changes in traffic flow.

  • Traffic lights at 6 intersections on Jinan Jingshi Road were optimized, reducing traffic delays by 10%

  • Traffic lights at 6 intersections on Jinan Weishier Road were optimized, reducing traffic delays by more than 15%

All this has just begun. Didi's current average daily order volume exceeds 20 million, but in the entire travel field, this number is not large, accounting for only about 1.5% of all travel.

From the perspective of inclusive travel, how to use cutting-edge technologies to do valuable things, how to make public transportation more stable and predictable, and how to accurately recommend the optimal transfer methods for multiple modes of transportation in real time, etc., these are all Didi's future focus areas.

Talking about dreams and feelings, Zhang Wensong said that from teaching at the university, to leaving to start a business, to Alibaba and now Didi, from a personal perspective, it feels better to make truly useful and valuable products, and everything he does will be from the perspective of solving problems for people and bringing value to society. In the process, he constantly makes choices on key issues and constantly solves problems, which is also a process of continuous learning and improvement for himself.

[[191941]]

[51CTO original article, please indicate the original author and source as 51CTO.com when reprinting on partner sites]

<<:  TensorFlow implements image completion based on deep learning

>>:  Exclusive interview with DeepMap COO Luo Wei: In the era of autonomous driving, how can startups break through in the field of high-precision maps?

Recommend

How to conduct an activity with a higher degree of completion?

How can an operator carry out an activity more sc...

Users (fans) start to leave at an accelerated rate, what else can you do?

If you are extremely sensitive to user activity, ...

To start running a WeChat store from scratch, just read this article

Many students who are engaged in operations shoul...

User operation: How to build a user portrait model?

At this stage, operators, no matter which positio...

7 key points you must know when starting to operate Douyin

I have summarized and sorted out some universal a...

Android Network--How I did it: Volley+OkHttp+Https

Use OkHttp as the transport layer implementation....

Xiaomi's 46 high-definition posters tell you how to do fan marketing

How did Xiaomi achieve rapid development in just ...

Worth collecting: 70 common mistakes in operations!

Introduction : Whether you are a newcomer or a ve...

How to evaluate whether KOL marketing is effective? Look at these 5 points!

In a recent survey, 76% of marketers said that th...

100 highly efficient tools, essential for operators and promoters!

Today I will share with you some common tools for...

75 Big Data Terms You Should Know

Part 1 (25 terms) If you’re new to big data, you ...

Operational tips: How to use new media to promote products

Before explaining new media operations, let’s fir...

How to plan group buying activities on mini programs?

The author was recently responsible for designing...