Flexible and efficient model-based congestion detection approach

Abstract

This paper addresses the problem of road traffic congestion detection. We propose an effective approach to detect traffic congestion by combining the piecewise switched linear traffic (PWSL) and Shewhart control scheme. This approach uses PWSL model to describe the evolution of traffic density, and Shewhart chart to detect traffic congestions based on residuals obtained from PWSL model. The PWSL-Shewhart approach is evaluated using traffic data from the four-lane State Route 60 (SR-60) freeway in California. Results indicate that our approach accomplished reliable detection of traffic congestion.

Anomaly detection and classification in traffic flow data from fluctuations in the flow–density relationship

“…Specifically, Zeroual et al (2017) looked at the residuals from such a model, smoothing them using an exponentially weighted moving average and then applied statistical tests to determine when to raise congestion alarms. The same model was considered in Abdelhafid et al (2018), where instead of a moving average, the authors considered upper and lower control limits for model residuals defined as three standard deviations away from some mean value. The authors again considered a similar approach in Harrou et al (2018), however using a generalized likelihood ratio test for determining if residuals belonged to one of two possible generating distributions.…”