Abstract
Travel time is one of the most critical indexes to describe urban traffic operating states. How to obtain accurate and robust travel time estimates, so as to facilitate to make traffic control decision-making for administrators and trip-planning for travelers, is an urgent issue of wide concern. This paper proposes a reliable estimation method of urban link travel time using multi-sensor data fusion. Utilizing the characteristic analysis of each individual traffic sensor data, we first extract link travel time from license plate recognition data, geomagnetic detector data and floating car data, respectively, and find that their distribution patterns are similar and follow logarithmic normal distribution. Then, a support degree algorithm based on similarity function and a credibility algorithm based on membership function are developed, aiming to overcome the conflicts among multi-sensor traffic data and the uncertainties of single-sensor traffic data. The reliable fusion weights for each type of traffic sensor data are further determined by integrating the corresponding support degree with credibility. A case study was conducted using real-world data from a link of Jingshi Road in Jinan, China and demonstrated that the proposed method can effectively improve the accuracy and reliability of link travel time estimations in urban road systems.
Highlights
Travel time is critical traffic information for road users and traffic managers [1]
The main contributions of this paper are as follows: first, a support degree algorithm among multi-sensor traffic data is proposed based on similarity function and log-normal distribution model, so as to solve the conflicts of different traffic data sources; second, a credibility algorithm of single-sensor traffic data based on membership function is developed to eliminate unrealistic erroneous data and exclude uncertainty of data detection, incorporating the effects of sample vehicle penetration rate; third, the reliable fusion weights of each type of traffic sensor data are determined by integrating the support degree with the credibility, which further achieves accurate and reliable link travel time estimates
On the basis of obtaining distribution functions of three link travel time series, we develop a support degree algorithm based on similarity function and a credibility algorithm based on membership function
Summary
Travel time is critical traffic information for road users and traffic managers [1]. It can better measure traffic congestion and transportation efficiency of urban roads, which is used as an indicator of traffic operating performance. Zhao et al adopted a gated recurrent unit model to predict travel time based on multi-source data [20] These methods can tackle complex data fusion problems, but have a higher requirement for the number of training samples. The main contributions of this paper are as follows: first, a support degree algorithm among multi-sensor traffic data is proposed based on similarity function and log-normal distribution model, so as to solve the conflicts of different traffic data sources; second, a credibility algorithm of single-sensor traffic data based on membership function is developed to eliminate unrealistic erroneous data and exclude uncertainty of data detection, incorporating the effects of sample vehicle penetration rate; third, the reliable fusion weights of each type of traffic sensor data are determined by integrating the support degree with the credibility, which further achieves accurate and reliable link travel time estimates.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call