Optimal Time Interval for Investigating Prior Information in Network Sensor Location Problem

Abstract

Based on various types of prior information, the network sensor location problem (NSLP) optimizes the locations of sensors in a network, providing comprehensive traffic flow information with a minimum number of sensors. However, few studies have considered the accuracy of prior information, and it is obvious that prior information inconsistent with reality will affect the results of NSLP. Generally, prior information can be obtained from other flow information. When solving NSLP, prior information requires additional or specialized field flow investigations to ensure its accuracy. In general, short-term prior information is adopted to infer the flow of information of interest. The purpose of this study is, therefore, to select the optimal time interval for investigating prior information. This study analyzed the types and characteristics of prior information, and proposed a mathematical method based on Pareto Optimality to identify the optimal time interval for investigating prior information, which can not only ensure the accuracy of prior information, but also reduce the investigation cost. Taking the turning ratio as an example, the traffic flow collected by sensors located at nine intersections on Zhongshan North Road in Shanghai, China was analyzed, and the turning ratios were calculated. Using the proposed method, the optimal time interval for investigating the turning ratio was determined to be 20 to 25 min. In addition, the most representative time intervals were identified as 9:00 to 10:00 a.m. It is recommended that investigating prior information in the early morning and in the morning and evening peaks be avoided.