Estimating Traffic Speed with Single Inductive Loop Event Data

Abstract

The estimation of traffic speed based on loop detector occupancy is a long-standing problem that is critical to traffic management and control. Single-loop detector stations are still popular in highway traffic systems in practice. Although dual loop stations could generate more accurate speed estimation with event data, updating single loops to dual loops is costly. Most previous work, such as the g-factor method on speed estimation with a single-loop system, used aggregated data and assumed a known effective vehicle length. However, those methods did not always produce good results, particularly in traffic transition phases. This paper proposes a new approach using event data and mode occupancy within a moving window of variable length to estimate the effective vehicle length, or g-factor. Berkeley Highway Laboratory dual loop data with a 1-Hz update rate and 60-Hz information were used for algorithm validation and reliability tests. The speed estimated from the corrected dual loop data was used as the ground truth for comparison. Root mean square error was used to quantify the discrepancy. Results showed that the proposed approach generates very satisfactory speed estimations compared with estimations from dual loop stations. This algorithm can be run at the control-cabinet level, in which event data of an individual loop from a 170 or 2070 controller are available without the need of any hardware modification. These controllers have the processor capabilities to handle data in real time.