交通需要予測諸手法の適用性の検討

概要

論文の詳細を見る
In order to clarify the usefulness of traffic demand predictors to be used in traffic management and control system, tested with actual field data were thirteen prediction algorithms consisting of UTCS-2, UTCS-3, methods of employing as estimates current measurements, historical means and measurements made a week before and models applying multi-variate regression analysis, time series analysis developed by Box and Jenkins, spectral analysis and Kalman filtering (made up of three models, Kalman-1, Kalman-2 and Kalman-3). The extensive tests were performed to see the applicability of the predictors for 5 to 60 min prediction of 5 min, 15 min and 30 min traffic demand. For the sake of ensuring the robustness of the predictors the tests were made for the case where the traffic demand of the day in question roughly follows the average historical pattern as well as the case where the deviation of the traffic demand of the study day from the historical average is substantial.The test results for two cases of prediction, 15 min prediction of 15 min demand and 30 min prediction of 30 min demand (which are most likely to be interesting from practical viewpoint), suggested that in the former prediction the Kalman-3 model (which employs, as state variables, not only traffic demand on the link in question but also on other relevant links, updating the transition matrix on a real-time basis) and the multi-variate regression analysis model (which also utilizes traffic demand informations on some links inclusive of the subject link, with parameters fixed throughout the study day) perform better than the remaining predictors while in the latter prediction only the Kalman-3 model yields the most reliable prediction algorithm.
社団法人電気学会の論文

著者

奥谷巖
信州大学

交通需要予測諸手法の適用性の検討

スポンサーリンク

概要

著者

関連論文

スポンサーリンク