详情

原文传递 Effects of Inaccurate Data on the Performance of Incident Detection Algorithms

题名：	Effects of Inaccurate Data on the Performance of Incident Detection Algorithms
作者：	Edmond Chin-Ping Chang, Ph.D., P.E.; Stephanie L. Kolb
关键词：	Effects; Inaccurate Data; Performance; Incident Detection Algorithms
摘要：	Incident detection algorithms are commonly used at Transportation Management Centers to automatically identify traffic patterns that may be indicative of an incident. These algorithms process 20 or 30 second real-time volume, occupancy, and/or speed data typically from loop detectors. However, due to the inaccuracies of loop detector data and the challenge of distinguishing between congestion and incident conditions, it is difficult to reliably detect incidents through algorithms. Using real-world traffic data collected from the Texas Department of Transportation (TxDOT) TransGuide system in San Antonio, the inaccuracies of loop detector data and the effects on algorithm performance were examined. Three algorithms were evaluated, including California algorithm #8 which many TMCs currently employ, the TxDOT Speed algorithm used at TransGuide, and California algorithm #8 with fuzzy logic. The paper also discusses the need for and means of filtering inaccurate data prior to applying data to incident detection algorithms. The examination of the loop detector data showed that occupancy data were commonly unreliable while speed data were usually reasonable. The study found that the speed-based algorithm worked well in detecting the seven incident cases that were evaluated, most likely because the speed data were more accurate. The performance results show that the fuzzy logic version of California algorithm #8 more effectively handled the inaccurate data and had fewer false alarms than California algorithm #8. To improve algorithm performance, filtering inaccurate data prior processing by algorithms is recommended.
总页数：	ITS America. Meeting (9th : 1999 : Washington, D.C..). New thinking in transportation : conference proceedings. 1999. pp19
报告类型：	科技报告