[Company news]AliCloud enables Zhejiang Communications Department to Forecast Traffic Jams with Big Data
Created#More Posted time:Dec 10, 2015 20:00 PM
AliCloud enables Zhejiang Communications Department to Forecast Traffic Jams with Big Data
"We will leave for your destination in 30 minutes. Please choose your route based on future traffic conditions." It seems that this question would beat any navigation software, as none of them knows what the traffic conditions of your destination will be in 30 minutes and whether the driver should make a detour or not.
Big data knows.
Zhejiang’s Communications Department is launching a new pilot program that will allow their software to combine historical traffic records with real-time highway data to accurately forecast traffic conditions within every next hour. The software, built on AliCloud's big data computing capabilities, can forecast traffic conditions to 91% accuracy.
According to Han Haihang, director of Zhejiang Traffic Information Center, forecasts for future traffic conditions enable Zhejiang Communications Department to improve traffic congestion and also enable drivers to plan their journeys better.
The AliCloud big data computing service (ODPS) provides analytical support for this program and there are several senior data scientists involved in its co-development. According to Dr. Min Wanli from AliCloud, ODPS' powerful computing capabilities make it possible to complete data analysis of historical
records for almost all of Zhejiang provinces 1,300km of highways in less than 20 minutes including providing real-time data analysis in less than 10 seconds.
Forecasting Real-time Traffic Conditions Based on Cell Phone Signals
To forecast the future, first you nee to understand the present. It has always been difficult to detect traffic conditions in real time. Due to the constraints of data collection technology, it usually takes a long time to update the real-time traffic data, sometimes more than 15 minutes. The traditional coil-laying method
requires a massive hardware investment.
For these reasons, Zhejiang Communications Department has introduced new technology to link phone signal data with traffic data. Generally, there is one cell-phone carrier station every 500m on city roads and every 2,000m on suburban highways. The signal data is collected when phone users pass by these base stations. As more phone signals are collected, the software can more accurately reflect the real-time traffic condition changes on a specific section of road during a specified
“For example, if we find that the sample vehicle stops on a highway or most of the vehicles on the same road section slow down, we can determine that an accident or traffic jam may have occurred on this road section.”
According to Han Haihang, compared to the traditional method which gathered highway traffic data using sensors, this method saves at least 90% in terms of cost and deployment is also shortened to only 2-3 months.
Self-Driving Cars: The Importance of Accuracy
On a rough estimate, by selecting appropriate route and departure times, drivers can reduce their travel time by 5% to 10% and lower their fuel costs by 2% to 10%.
Forecasts of future traffic conditions can also be used to support the currently being developed self-driving car technology. Besides using various sensors to quickly analyze current traffic data, self-driving cars will also need to know the traffic conditions on
upcoming sections of road in order to accurately guide the care to most convenient route.
Min Wanli says that “traffic conditions forecast is of high application value. However, more attention needs to be placed on improving its accuracy, speed and affordability. Simple forecasts based only on historical averages will not have any practical meaning. The more analysis factors and dimensions we use, the richer data we gather and the more accurate the forecast result will be.”
Microsoft partnered with a Brazilian university earlier this year in a similar test which produced an accuracy of 80%.
"Road network relationships, upstream/downstream events, and even other external comprehensive factors such as weather conditions should be included. But once these massive volumes of data are incorporated into the spatiotemporal evolutionary model of the entire road network, the cloud platform's big data processing capabilities need to be improved to handle this demand." According to Min Wanli, “AliCloud is the world leader in this aspect.”
It is reported that AliCloud sorted 100TB of data in less than 7 minutes (377 seconds) in the Sort Benchmark competition, setting a world record. Sort Benchmark is considered the Olympics in the big data field. Each year, top companies and academic institutions all over the world would participate in this competition to have their latest research results evaluated.
[Cloudy edited the post at Dec 15, 2015 14:23 PM]
1st Reply#Posted time:Jan 6, 2016 17:07 PM