AsiaIndustrial NetNews: On March 27, Ye Jieping, Vice President of Didi Chuxing Research Institute, attended the 2017 Xinzhiyuan Open Source Ecological AI Technology Summit. He also said that artificial intelligence technology has been widely used in all aspects of Didi, which can continuously improve user travel efficiency and optimize travel experience. Didi’s layout and exploration in the field of big data and artificial intelligence has been at the forefront of the Internet industry.
Ye Jieping pointed out that Didi Aiming is a world-class travel problem, and the calculation behind it is actually far more complicated than AlphaGo. Whether it is vehicle scheduling or driver-passenger matching, the dimension, complexity and real-time nature it considers far exceed other industries. At present, Didi is using machine learning and big data to solve problems such as intelligent scheduling and supply and demand forecasting. Through the continuous optimization of the model, the computing time is shortened, and the forecasting effect is improved at the same time, making people’s travel more convenient and experience better. “Didi is a data, technology and experience-driven company, whether it is predicting the destination before the passenger issues the order, recommending the pick-up point, or intelligently dispatching the order after the order is issued, route planning, and even safe driving during the itinerary, itinerary After the end of the driver and passenger judging link, artificial intelligence technology has been widely used, and artificial intelligence has become the core of Didi technology,” Ye Jieping said.
In his speech, Ye Jieping also gave a detailed explanation of Didi’s intelligent dispatch. Intelligent dispatch is one of Didi’s core technologies. Every time a passenger issues an order, it is necessary to use large-scale distributed computing to optimally match drivers and passengers. In order to maximize the platform efficiency and user experience, the optimal driving path should be calculated to minimize the total time.
However, unlike the static stop of goods, information and other information when searching online, the vehicle is always moving, and the driver may pass an intersection or drive to the expressway after a few seconds. This also places higher demands on matching—the need to be able to predict future conditions and quickly perform dynamic, real-time matching of drivers and passengers.
(Ye Jieping, vice president of Didi Chuxing Research Institute, explains the path planning algorithm in detail)
Two map technologies, route planning and ETA (estimated time of arrival), are the keys to Didi’s optimal matching. According to Ye Jieping, through mining and learning from the massive user driving data of Didi Chuxing, Didi has designed a brand-new intelligent path planning algorithm around the lowest price, highest driver efficiency and optimal transportation system operation efficiency, which can help predict the future. Make accurate predictions of road conditions, consider all possible future movements of the driver as a whole, and calculate the optimal path from point A to point B in milliseconds.
As early as 2015, Didi applied machine learning to ETA for the first time nationwide based on massive travel data. By using a new time estimation algorithm and feature mining, it greatly improved the accuracy of time estimation. . On this basis, DiDi has recently innovatively applied deep learning to ETA, further improving the prediction accuracy.
Big data and artificial intelligence technologies have also driven Didi to further improve the efficiency of dispatching orders. Ye Jieping revealed that before Didi can make a global judgment every 2 seconds, in the rapid large-scale calculation, it can complete the global optimal intelligent dispatch. Now this dispatching algorithm has been upgraded, it can take into account the overall efficiency of drivers within a day based on the prediction of supply and demand and travel behavior throughout the day. Maximize revenue.”
Behind the development and application of these big data and artificial intelligence, it also reflects Didi’s powerful cloud computing capabilities. According to public information, Didi completes more than 20 million orders per day, processes more than 2,000 TB of data per day, which is equivalent to 2 million movies, and has more than 9 billion daily route planning, which is equivalent to about 6 million times per minute. At present, Didi is driving the rapid iterative upgrade of artificial intelligence technology, and has also built an intelligent system, Didi Brain, which can maximize the use of transportation capacity through big data, machine learning and cloud computing, and make optimal decisions for everyone. Users design the most intimate and intelligent travel plan. In the future, Didi will also actively join hands with city managers to build a smart transportation system and create a new ecosystem for future travel.
The Links: 3HAC022194-001 SGDR-AXA02A DISPLAY