Urban Expressway Incident Detection Algorithm Based on Floating Car Data

Abstract

Easily achieved automatic incident detection (AID) algorithm based on two-dimensional integration was proposed by the analysis of floating car data (FCD) and traffic event data. The wide coverage of FCD was applied to AID of urban expressway to improve the performance of incident detection. The algorithm integrated the change of speed by the standard difference in the temporal dimension and the difference between travel speed on upstream and downstream in the spatial dimension to detect incidents. Then, the algorithm was verified with the FCD and traffic incident data collected from Beijing expressway, with a 87% detector rate (DR) and a 1.86% false alarm rate (FAR), which were extremely close to the average TMC acceptable limits of at least 88% of DR and not more than 1.8% of FAR. Therefore, we could conclude that the FCD based integrated algorithm could meet the basic needs for practical applications. 1. BACKGROUND As a basis of the main road network, the urban expressway plays a vital role in the urban road network. If an accident occurs, it might lead to secondary accidents which would aggravate congestion on the urban expressway congestion, if they are not settled in time due to the presence of high-speed traffic. Therefore, accurate and rapid AID has a great significance in reducing the adverse effects of incidents on the urban expressway traffic. At the same time, AID is one of the key functions of traffic management and control system. Detection and confirmation of the time, location and type of the incident accurately and timely are the basis to improve the reaction ability of traffic emergency management. The performance of AID algorithm is the core of traffic incident management, and also an important evaluation index of successful operation in intelligent transportation systems. Traffic incidents can be detected through various means such as manual observation, detection algorithm, mobile phone, etc. However, the most effective method is detection algorithms, which can automatically detect incidents through data analysis and evaluation, and data source can be diverse such as fixed detectors, video detectors and FCD at certain time intervals. In former research, most AID algorithms were successfully developed by using loop-detector-based traffic measurements. But its shortcomings were