Technical Seminar on Traffic Signal Optimization with Crowdsourcing Vehicle Trajectory Data

By Ir Shirley LEUNG

While the parameters of traffic signal system are normally adjusted every 2 to 5 years in US and about every 6 months in the Mainland China with regard to the latest traffic situation, would it be possible for the system to be adjusted in real-time?  Dr Henry Liu, the Chief Scientist of DIDI Smart Transportation of China, in this technical seminar held on 24 October 2017 in the University of Hong Kong told us that it was possible, and indeed had been implemented in 8 cities in China like Jinan, Shenzhen and Beijing. 

Traditionally, the traffic signal parameters are optimized either by manually collected data or data from sensors such as loop or video detectors which are costly.  With cloud computing, machine learning, and the world’s largest mobility service platform which continuously keeps track of over 20 million orders a day, DIDI is able to estimate the real-time traffic situation and perform real-time optimization of traffic signals by making use of their massive volume of vehicle trajectory data collected from their DIDI vehicles.  DIDI has been proactively collaborating with the city government free-of-charge and making use of their vehicle trajectory data to reduce traffic congestion.  Every week, they carry out the traffic signal optimization and send the updated parameters to the Police for their implementation, and the results are fascinating.  For example, the use of DIDI’s vehicle trajectory data has successfully resolved the traffic gridlock problem on the most congested roads in Jinan, with traffic delay reduced by more than 20% and traffic speed increased by more than 25% during peak hours! 

Why DIDI as an enterprise has such an enthusiasm to help the government for free?  Dr Liu explained that it was a win-win solution, as by reducing the traffic congestion in the city, the service of the DIDI vehicles could be enhanced and their business revenue could be increased.

The seminar was insightful and was well appreciated by all the around 40 participants.   Because of the insufficient vehicle trajectory data available in Hong Kong, I believe this innovative approach for traffic signal optimization is unlikely to be adopted in Hong Kong in the near future.