This paper describes work in progress on the development of practical and accurate forecasting methods for the DYNA project for developing, implementing and testing a real-time traffic model for inter-urban motorway networks. It was decided to apply the following methods in parallel, and merge their results: (1) a statistical model, using a state space/Kalman filter approach with explicit feedback of prediction errors; and (2) a dynamic traffic assignment model, working with a dynamic origin-destination (OD) prediction method. Model (1) was expected to be most accurate for between one and 15 minutes ahead, while model (2) was expected to be more accurate for between 15 and 60 minutes ahead. The paper ahead. The paper discusses: (1) the proposed system architecture; (2) OD estimation and prediction; (3) dynamic traffic assignment; (4) the component parts of the statistical model; and (5) the merging of very short and short-term traffic predictions. So far, the project has specified the algorithms and designed the software in detail, and tested the models against actual measurements. The present model structure is provisional, and indicates a framework for development of the eventual final model.
Samenvatting