• PDF version [PDF: 908KB]

East Nippon Expressway Co., Ltd., Ltd. (hereinafter, NEXCO EAST) and NTT Docomo Co., Ltd. (hereinafter, docomo) have renewed "AI Congestion Prediction" ^{* 1} under demonstration experiment at Tokyo Wan Aqua-Line Expressway (hereinafter, Aqua-Line). We will start providing the time required to pass the Aqua-Line by band (hereinafter referred to as the required time ^{* 2}) and traffic demand ^{* 3} from Saturday, December 22, 2018.

Aqualine is a real-time demographic created using the mobile phone network mechanism. ^※Four (The following is demographic statistics) and past traffic congestion records and regulation information held by NEXCO EAST are used to predict traffic congestion at the time of returning home by "AI traffic congestion prediction" developed by DoCoMo using artificial intelligence (AI) technology. Since December 2017, we have been conducting a field trial. Through experiments, customers are provided with information such as congestion start times and peak congestion distances on the NEXCO EAST customer website “DraPla”.

This renewal is the first in the world to predict the required time and traffic demand for Expressway from the number of people on the day to several hours ahead, as determined by demographics. ^※Five It is possible to tell you the required transit time every 30 minutes, etc.

After verification of the demonstration test, both companies will proceed with studies for full-scale introduction and expansion to other routes.

According to the results of the customer survey of the "AI traffic jam prediction" demonstration experiment that has been underway since December 2017, about 90% of all customers, especially those who frequently use aqualine ^{* 6} In this case, about 95% of the respondents indicated that they would like to continue using "AI traffic jam prediction" when using the aqua line. Also, as future expectations for "AI traffic congestion prediction", we were able to confirm strong requests regarding information provision for each time zone and provision of required time.

We have developed this technology based on these requests. The accuracy of traffic congestion prediction is greatly improved by this technology, and the ratio of the time zone in which the required time error is 30 minutes or more in the time zone predicted to have traffic congestion is ^{* 7} The ratio of 0.8% to 8.2%, the ratio of 20 minutes or more is 6.7% to 26%, and the error of the forecasted traffic demand is 9% (average 16 cars per 10 minutes) against the actual traffic volume. Was ^{* 8} .

• Target of evaluation: January 1, 2015 (Thursday) to September 30, 2018 (Sunday) (excluding accidents and regulations)
• The time zone when there is traffic congestion is the time zone in which it is predicted that the required time will increase corresponding to traffic congestion of 1 km or more (assuming a traffic speed of 20 km)
• AI traffic congestion prediction up to now is calculated based on traffic congestion start time, peak time, traffic congestion cancellation time, maximum traffic congestion length

NEXCO EAST is developing a "safe, secure, comfortable and convenient Expressway service" in its medium-term management plan. In addition, in the medium-term strategy 2020 “beyond declaration,” DOCOMO is working with its partners to resolve social issues. The two companies will continue to study the further utilization of "AI traffic jam prediction" and its expansion to other routes, and will work to further strengthen traffic problems.

1. AI traffic jam prediction is a technology that constitutes the NTT Group's AI "corevo®".
2. Time required to pass from Kisarazu JCT to Kawasaki Ukishima JCT
3. Traffic demand is the number of vehicles that potentially pass through the Expressway at each time of day, and corresponds to the amount of traffic when there is no limit to the amount of traffic (traffic capacity) that can flow on the road.
4. The demographics used in this experiment are accelerated versions of DoCoMo's "mobile spatial statistics." This is information that indicates the number of people in each group by area and attribute, and cannot identify individual customers. The demographics used in this experiment comply with the "Mobile Spatial Statistics Guidelines" that summarizes the basic items when creating and providing mobile spatial statistics in order to strictly protect the privacy of customers.
   ・Mobile spatial statistics guidelines
   ( https://www.nttdocomo.co.jp/corporate/disclosure/mobile_spatial_statistics/guideline/index.html )
5. DoCoMo research on Friday, December 21, 2018
6. Customers living in the Kanto region excluding Chiba prefecture who use Aqualine at least once every six months for leisure purposes on holidays
7. What is a congestion forecast calendar? A congestion forecast created and published by NEXCO EAST on past congestion records.
8. Comparison with non-congested traffic volume that is not restricted by traffic capacity

• "Corevo®" is a registered trademark of Nippon Telegraph and Telephone Corporation. ( http://www.ntt.co.jp/corevo/ )

Attachment

• Attachment 1: Overview of new technologies [PDF: 412KB]
• Attachment 2: Overview of customer survey results [PDF: 212KB]
• Attachment 3: About forecast traffic demand [PDF: 399KB]
• Reference: Overview of each company [PDF: 99KB]