Crowdsourcing Phase and Timing of Pre-Timed Traffic Signals in the Presence of Queues: Algorithms and Back-End System Architecture

This paper describes a crowdsourcing-based system for phase and timing estimation of pre-timed traffic signals. The input crowd is a real-time feed of sparse and low-frequency probe vehicle data, and the output is an estimated collection of Signal Phase and Timing (SPaT) information. The estimations could be ultimately fed into a connected vehicle's driver assistant application. Different from the authors' previous work, the approach described in this paper ensures the accuracy of the SPaT estimations even in the presence of queues. This was achieved by investigating the probe data influenced by the heavy traffic and the delay in queues. This paper is also a sequel to the authors' previous work as it provides an in-depth overview of the crowdsourcing algorithms and their back-end implementation. The accuracy of the crowdsourcing algorithm is also experimentally evaluated for a selection of pre-timed traffic lights in San Francisco, CA, USA, by utilizing a real-time data feed of San Francisco's public buses as an example data source.