2017/02/03 by Shumeet Baluja, Baluja, Shumeet, Michele Covell +3 · 1 voice
Computer Science · Engineering · Social Sciences · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Systems and Control (eess.SY) #Traffic Prediction and Management Techniques #Traffic control and management #Transportation Planning and Optimization #cs.AI #cs.LG #eess.SY #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1702.01205
openalex publication_date 2017/02/03 · arxiv published 2017/02/03 · arxiv updated 2017/02/03 · openalex created_date 2022/08/28 · openalex updated_date 2026/07/28
Real-time optimization of traffic flow addresses important practical\nproblems: reducing a driver's wasted time, improving city-wide efficiency,\nreducing gas emissions and improving air quality. Much of the current research\nin traffic-light optimization relies on extending the capabilities of traffic\nlights to either communicate with each other or communicate with vehicles.\nHowever, before such capabilities become ubiquitous, opportunities exist to\nimprove traffic lights by being more responsive to current traffic situations\nwithin the current, already deployed, infrastructure. In this paper, we\nintroduce a traffic light controller that employs bidding within micro-auctions\nto efficiently incorporate traffic sensor information; no other outside sources\nof information are assumed. We train and test traffic light controllers on\nlarge-scale data collected from opted-in Android cell-phone users over a period\nof several months in Mountain View, California and the River North neighborhood\nof Chicago, Illinois. The learned auction-based controllers surpass (in both\nthe relevant metrics of road-capacity and mean travel time) the currently\ndeployed lights, optimized static-program lights, and longer-term planning\napproaches, in both cities, measured using real user driving data.\n