New research identifies curves, investigates how drivers react to them

Negotiating curves – whether traveling along a bend in the road or going up a hill – is one of the more complex maneuvers for a driver. A number of factors may contribute to the level of complexity including presence of other vehicles, lighting conditions, time of day, absence of paved shoulders and roadside conditions. We, as drivers, make decisions as we enter, drive through and exit curves about our speed, braking and choice of path. Researchers are interested in the relationship between these driver decisions and the characteristics of the curves. Understanding those relationships is the first step toward developing countermeasures that will aid drivers in making better decisions.

One of the data gaps that exists in many inventory databases is the location and characteristics of vertical curves (hills and valleys) in the roadway system. As one of the first groups to receive funding from the Strategic Highway Research Program (SHRP2) Implementation Assistance Program, HSRC was tasked with determining how to use SHRP2 data to identify the location of the vertical curves and the features of these curves. The ultimate goal of the project was to develop a new tool for the acquisition of this type of data for safety data analysis.

Here’s what happened: SHRP2’s massive data compilation consists of detailed roadway data, driver behavior information and vehicle kinematics. The roadway inventory database contains grade data (percentage incline or decline of the roadway) measured at approximately every 25 feet along a road. This information, along with numerous other data elements, were collected using an instrumented vehicle that drove the roadways. From this grade data, HSRC researchers employed a “sliding window” analysis to extract the vertical curve information. By incrementally sliding a window of a fixed distance along a route, the team was able to determine if a vertical curve was encountered by identifying inflection points where there was a change in grade.

With a vertical curve identified, the next question was how the accuracy of the curve parameters can be assessed. The HSRC team compared two methods, each assessed against two scenarios: 1) a hypothetical road with artificially created vertical curves, and 2) a section of an actual road in Washington State. The best performing method used a linear correlation element to detect when a set of road segments formed a vertical curve. This method accurately identified the presence of vertical curves without falsely identifying curves where none existed, although it generally underplayed the length of the curves.

The development of this new linear correlation method is an improvement over analyzing the simple difference of average grades, which was the previously accepted practice. Comparisons of a hypothetical road to a real-world road proved useful to identify the best performing method and to determine the optimal parameters for the most accurate results.

The development of this vertical curve identification tool supported a larger goal. The second phase of the study combined these identified vertical curves with horizontal curve data and investigated driver behavior – particularly speed and lane deviation – on various combinations of alignment. Results showed that drivers approaching a horizontal curve (or hill) have the worst performance in terms of lane deviation. In addition, sharper horizontal curves (meaning curves on the road from left to right) are associated with higher likelihoods of lane deviation.

Using these new analysis tools and methods, researchers can now identify certain segments of a road where drivers may be more susceptible to risk. And this knowledge can be used to prioritize potential improvements in the design and operation of a particular roadway that will improve safety.

Executive Editor: Caroline Mozingo
Managing Editor: Patty Harrison
Graphic Designer: Graham Russell