Enhancement of Automatic Incident Detection Algorithms for Singapore's Central Expressway

Abstract

Timely detection of accidents, vehicle breakdowns, and events that obstruct normal traffic flow is critical to successful implementation of an incident management system in combating traffic congestion along expressways. The purpose of this study is to enhance the performance of existing expressway automatic incident detection algorithms in detecting lane-blocking incidents. In this study, a video-based vehicle detector system was used along the Central Expressway (CTE) in Singapore to collect 160 incidents. Two main tasks were carried out with the CTE incident database to investigate factors that influence incident detection performance and to investigate the use of these findings to enhance existing CTE-calibrated incident detection algorithms. Results indicated that the inclusion of preincident traffic flow or occupancy conditions and the use of traffic speed together with occupancy in an algorithm would yield enhanced detection performance. Of the algorithms studied, the dual variable algorithm, which uses traffic speed and occupancy, can consistently give the best detection performance. From an efficiency perspective, there were no significant changes in time lag in the detection of an incident. A comparative evaluation suggested that the occupancy-based algorithms were generally more effective than the flow-based algorithms in detecting incidents.