The single-channel detection scheme has been widely used in practice for traffic control. Differing from lane-by-lane detection, single-channel detection has all loop detectors across multiple lanes wired together and provides a single input to the controller. Although single-channel (wiredtogether) detectors are commonly used by traffic agencies, the traffic volume data collected from these single-channel loops are inaccurate because of high misdetection rates, compared with lane-by-lane detection. This study focuses on developing a probability-based approach for correcting the traffic volume data collected by the single-channel loop detectors at signalized intersections. The proposed probability-based nonlinear model (NM) explicitly describes a potential model compared with a multiple linear regression model. Both models were calibrated and validated using real-life data from seven two-lane intersection approaches and three threelane intersection approaches under various traffic conditions. The results showed that the proposed NM requires much less effort to calibrate and can effectively reduce system errors by reducing absolute mean errors by 60% on average. The correction effect of the model positively scales with increases in the system error rate. The verification results indicate the proposed NMs are capable of correcting the misdetection resulting from single-channel loop detectors and accurately estimate hourly traffic volume. The NM demonstrated its transferability to different locations and can be a general application in practice.
ASJC Scopus subject areas
- Civil and Structural Engineering
- Mechanical Engineering