Data-Driven Approaches for Robust Signal Plans in Urban Transportation Networks
- Author(s): Amini, Zahra
- Advisor(s): Skabardonis, Alexander
- et al.
In urban transportation networks with signalized intersections a robust pre-timed signal plan is a practical alternative to adaptive control strategies, since it has less complexity and an easier implementation process. Recent advances in technology are making data collection at traffic signals economical and data-driven approaches are likely to benefit from the large traffic data. Data-driven approaches are necessary for designing robust timing plans that can satisfy rapid traffic volume fluctuation and demand growth. This dissertation introduces four data-driven approaches for studying and improving traffic conditions at signalized intersections.
Firstly, I discuss the development and testing of two algorithms for checking the quality of traffic data and for estimating performance measures at intersections. The first of these algorithms estimates the systematic error of the detector data at signalized intersections by using flow conservation. According to the ground truth data from a real-world network, the algorithm can reduce the error in the data up to 25%. The second algorithm helps in estimating intersection performance measures in real-time by measuring the number of the vehicles in each approach using high resolution(HR) data.
An offset optimization algorithm was developed to adjust signal offsets so as to improve the delay in the system. The performance of three real-world networks using the offsets obtained by the algorithm and those obtained from the widely used Synchro optimization tool, are compared using the VISSIM microscopic simulation model. Simulation results show up to a 30% reduction in the average number of stops and total delay that vehicles experience along the major routes when using the proposed algorithms’ optimized offsets. The fourth algorithm estimates the appropriate switching time between designed timing plans during the day based on the traffic profile of the intersection by using the K-means clustering method.
In conclusion, these four algorithms extract useful information from HR data about traffic at signalized intersections. Moreover, the algorithms assist in designing robust timing plans for satisfying demand fluctuations at signalized intersection. Lastly, simulation results from real-world networks illustrate the significant improvements that the proposed data-driven approaches can make in the control systems at urban transportation networks.