Modeling and evaluation of incident detection algorithms

Abstract

Traffic congestion is a growing problem for Istanbul. The application of Intelligent Transportation System (ITS) is aimed at reducing traffic delays due to recurrent and non-recurrent congestions. The non-recurrent congestions such as incidents, vehicle breakdowns or other unexpected events need timely attention to control the traffic and to turn back to its usual condition. For this reason, incident detection can be seen as a significant tool in minimizing the negative effects of incidents. Timely and accurate incident detection is implemented through the automated incident detection (AID) algorithms, which detect incidents from traffic data. This study aims to investigate the performance of several existing incident detection algorithms on the D-100 Highway. For preliminary analysis, the California, Minnesota, Standard Normal Deviate (SND) algorithms are tested using the BUTSIM (Bogazici University Traffic Simulator) two-lane microscopic traffic flow simulator. While the test results for California and Minnesota are found acceptable, the SND algorithm and Kohnert Principle does not give efficient performance. The selected part of the D-100 Highway is modeled using PARAMICS, traffic microsimulation software. The California, Minnesota and SND algorithms are also tested in this model using API files created in another study. The California and Minnesota algorithms give an efficient performance. Furthermore, Lane Control Signs (LCS) are analyzed which are employed by TCC (Traffic Control Center) and ISBAK A.S. (Istanbul Transportation, Communication and Security Technologies Co.). The system is implemented to control the speed variations in D-100 Highway and inform drivers about the road conditions such as weather or incident. The effectiveness of LCS system under single-blocked-lane incident is evaluated through PARAMICS simulation experiments. It is found that the LCS system can reduces the travel times and increases link speeds with an appropriate compliance rate.