Evaluation of travel time estimation techniques in 'DACCORD'

摘要

The subject of this paper is the evaluation of travel time estimation techniques implemented in the DACCORD project. The DACCORD project is concerned with the development and application of co-ordinated traffic control in corridors and networks. A common feature of dynamic traffic management tools is their dependency on on-line estimates or predictions of travel time, based on inductive loop measurements.The objective of this study is evaluate the performance of the travel time techniques implemented in DACCORD, both with respect to the true travel time and each other. Special attention was paid to the comparison between theoretical performance and actual performance.The travel time estimators implemented in DACCORD are the instantaneous Network-level Travel time (NLTT), the "weighted" instantaneous NLTT, the off4ine dynamic NLTT all based on Section-level travel times (SLTT's) calculated from traffic speed, and theinstantaneous NLTT calculated using a vehicle conservation approach (the "mass balance algorithm").Additionally results obtained by an independent simulation study with the microscopic simulation model FOSIM are reported, which show a time lag for all estimators, except for the dynamic NLTT estimators. All techniques are accurate under free-flowconditions, but underestimate with the building up of congestion. Techniques behave differently under the resolution of congestionAn empirical study at the Amsterdam orbital motorway(A10) was carried out, in which reference observations of true travel time were made with a license-plate survey. These observations were compared with the travel time estimates using visual means and aquality indicator. The empirical results indicate a small difference between the instantaneous and dynamic travel time. With congestion under estimation of the observations is the result.The difference between the instantaneous and dynamic travel time indicate with the simulation study is not present with the empirical study in the same extent. The instantaneous is a relatively good estimator for on-line usage. Relative to the theoretical best travel time estimator only small improvements can be made. Other estimation techniques are more complex of algorithm, but do not improve the quality of the estimates.