Estimation of Multiclass and Multilane Counts from Aggregate Loop Detector Data

Abstract

Lane utilization on the highway is affected subtly by dynamic traffic management systems such as speed controls and lane management. To optimize the operation of dynamic traffic management, a better understanding of lane utilization is required, in particular, of how the flows of different vehicle classes (e.g., passenger cars, lorries) vary across the carriageway. Most loop detector systems do not collect this multilane, multiclass count data. This study developed a procedure for estimating multilane, multiclass counts from a variety of standard aggregate loop data formats from around the world. The estimation procedure involved the inference of multilinear regression laws that relate multilane, multiclass data to standard aggregate formats. The regression laws were then trained with small samples of individual vehicle data on a site-by-site basis. Preliminary results showed that the estimation procedure worked rather well, even when the input data were minimal—the extreme case being that of (U.S.-style) single-loop data, for which only flow and occupancy were available on a by lane basis. An error analysis indicated that small amounts of individual vehicle data were sufficient to train the estimator, provided they contained a representative mix of the flow behaviors at the site in question. Further work is required for the practical development of the tool, but it appears to have a wide range of potential uses for both researchers and practitioners.