In this work we build on the idea of using Bluetooth® sensors as a new intelligent transportation system application of estimating travel time along a section of a highway. Given the existence of High Occupancy Vehicle (HOV) lanes and Express lanes in the U.S highway network, a mixed population estimation problem naturally arises. This estimation problem is attacked from three dierent perspectives: (i) in light of the Expectation Maximization (EM) algorithm, (ii) using Maximum Likelihood Estimation (MLE) techniques and nally (iii) applying a cluster-separation approach to our mixed dataset.

The robust performance of the rst approach leads to an EM-inspired MLE technique, a hybrid of (i) and (ii) which combines the good estimation accuracy of EM based algorithms and the lower complexity of MLE techniques. The limitations and performance of all four approaches are tested on actual vehicular data on different highway segments in two dierent U.S states. The superiority of the hybrid approach is shown along with it's limitations.