|Institution:||Delft University of Technology|
|Keywords:||driving behaviour; work zone; unmanned helicopter; vehicle detection; vehicle tracking|
|Full text PDF:||http://resolver.tudelft.nl/uuid:9cf91ce4-c367-48f4-b612-7f5cbf5e70be|
Problem definition and objective. The performance of work zones with specific layouts cannot be accurately determined, because changes in individual driving behaviour at work zones are unknown. Therefore, the research objective is: “Gain insight into possible adaptation effects related to the layout of work zones, in empirical individual driving behaviour and subsequent in macroscopic effects, by presenting the applied method”. Unmanned helicopter and observed layouts. Empirical individual driver data have been collected using an innovative method: the unmanned helicopter instead of a manned helicopter (to reduce costs and collect more data). The three observed configurations vary in lane widths, channelizing devices, and presence of reverse curves. Data processing. To analyse individual driving behaviour, vehicle movements have to be extracted from stabilized frames, but difficulties appeared by stabilizing imagery and detecting and tracking vehicles. Therefore, new methods are introduced to detect vehicles: • ‘Edge detection’; • A program identifying mouse clicks locations; Also a method is implemented for tracking vehicles: • A shortest-path algorithm: the Dijkstra algorithm. Data analysis and results. For each configuration, empirical variables, e.g. time headway distributions, and the speed - following distance relationship are analysed and compared with consistent results: Driving speeds at corresponding following distances decrease by applying a work zone (and even more when also applying a reverse curve). Macroscopic effects are analysed by performing simulations, using a calibrated car-following model (the extended IDM) with individually -on their definition- estimated model parameters: • Smaller lane widths and a barrier as channelizing device reduced capacity with 1.5%; • When also applying a reverse curve, the reduction was even 6.2%.