Abstract

Traffic light control is one of crucial issues in the traffic deployment. However, the current procedure in generating traffic light schedule that relies on either empirical law manually or simulating software is time-consuming and lack of both flexibility and immediacy. Given real-time traffic flow information from vehicle license plate recognition sensors, the proposed framework can immediately generate a traffic light schedule to relieve arising or probable traffic congestion which was not expected at first by management authorities. The novelty of our system is that it combines the instant traffic flow transition estimation and an adaptation of Gaussian Mixture Model. Preliminary evaluations show that proposed method outperforms the current procedure that the traffic management authority in Tainan City takes.

Date

2020

Authors

Fandel Lin, Shiuan-Tyng Lin, Jie-Yu Fang, Hsun-Ping Hsieh

Book

Companion Proceedings of the Web Conference 2020

Pages

100-102