In the road design process, speed variation along the road segment is an important issue to consider in adapting road geometry to drivers' expectations. To achieve this objective, speed criteria are used to evaluate road consistency. Being able to estimate the operating speed in the design phase can lead to safer road alignment. With this objective, several researchers have developed operating speed models. Most of these models are based on collected spot speed data. They assume constant speed on curves and, therefore, deceleration that occurs entirely on the approach tangent. According to these assumptions, spot speed data are collected at the center of the horizontal curve and at the midpoint of the preceding tangent to obtain operating speed models. This paper presents a new methodology based on the use of Global Positioning System devices that allow continuous collecting and processing of speed data. With this new methodology, not only can new and more accurate operating speed models be developed, but cited hypotheses can also be checked. Observed speed continuous profiles allow studies that previously could not be done, especially as related to deceleration and speed variations. This study calibrated new speed models, including three for horizontal curves with a radius curve and the curvature change rate of a single curve as explanatory variables, and one for tangents that incorporates the curve speed model. Tangent-curve speed variations are evaluated, with comparison of Δ₈₅V and ΔV₈₅, analysis of the deceleration length occurring on a curve, and development of two deceleration models.

Drivers need sufficient passing sight distance (PSD) to pass slower vehicles with safety. This distance can help to improve traffic operation on two-way, two-lane highways. Existing models propose different values of PSD because of different assumptions. In only some cases were these models based on field data of passing maneuvers. This research proposed the design of a methodology to observe passing maneuvers on existing highways with the help of six video cameras installed at a fixed point next to passing sections. The use of more cameras allows complete registration of trajectories along the entire passing zone, with uniform image resolution. The methodology was applied to register a sample of 234 maneuvers on four passing zones. Trajectories of 58 maneuvers were completely described and analyzed with specific restitution software. Results were compared with those from existing PSD models. The distances traveled proposed by the AASHTO model on the left lane were (a) similar to average observed distances when the passed vehicle was one truck and (b) between 50 and 100 m higher when one passenger car was passed. Higher differences, greater than 100 m, were found between measured data and the PSD model (published previously), especially at high design speeds. The observed average speed difference between passing and impeding vehicles was significantly higher than that in any model. Variables with the strongest influence on the time and distance traveled on the opposing lane were the type and speed of the passed vehicle and the length of the passing zone. Left-lane time and distance increase with this length.