Real-time Traffic Estimation for Advanced Traffic Information in Berlin



Real-time Traffic Estimation for Advanced Traffic Information in Berlin

Authors

M Fellendorf, P Vortisch, PTV AG, DE

Description

Abstract

Berlin is Germany?s capital and cultural center with a population of about 4 Million people. Like many other metropolitan areas Berlin is facing traffic problems which should be partly relieved by managing transport by better traffic information for all travelers. As such the state of Berlin awarded a consortium comprising DaimlerChrysler Services AG and Siemens AG a contract at the end of 1999 to set up and operate a traffic management center called VerkehrsManagementZentrale (VMZ) Berlin. In April 2001, the two partners founded the joint operating company, VMZ Berlin Betreibergesellschaft mbH. This joint venture will operate VMZ Berlin over a period of 10 years. VMZ was set up as a public-private partnership with the investment costs for hardware and software covered by the Berlin government. After 10 years of operation the system will be owned by the Berlin government. The operating cost are split by the government and the private consortium. The traffic management center had been started to be developed in fall 2000 and will be finalized by summer 2003.
The private consortium has to provide basic mobility services for free supplemented by value added services for a fee. VMZ free services contain traffic information like current roadwork, parking situation, maps displaying congestion, life webcam pictures and route planners for private and public transport. This mobility information can be accessed for free under http://www.vmzberlin.de. Value added information like customized travel planning will be displayed via SMS or other means like upcoming smart phones. It will depend on the creativity of the consortium and market response as to which services will be available for a fee only. Hotels are already ordering location maps with access and routing information from the VMZ operating company. Other commercial services will include the option to pre-book car parking spaces over the Internet and alternative on-trip route guidance to avoid congestion.
It is not intended that the traffic management center will take an active role in controlling traffic by traffic lights or motorway control. The traffic management center will gather data which will be used for the future Berlin traffic control center. Under the liaison between public agencies of the State of Berlin and the private consortium, information and recommendations for all travelers will be linked with traffic monitoring and control. The data is gathered in the traffic management center and can be used by the traffic control center. Traffic control authorities will have access to comprehensive maps of the current traffic situation and traffic forecast which can be used for decisions on traffic control. Since the management center is operated by a private consortium it can not be in charge for active operation of traffic lights and mandatory signs. However it is expected that travelers will adjust their travel behavior such as mode choice, departure time choice and on-trip alternative route choice according to information provided by the management center. As such the center will provide implicit traffic control by traffic advice rather than explicit traffic control which is still under public responsibility. VMZ Berlin gathers all the important information relating to the traffic situation in the city. Static and dynamic data of public and private transport is already collected in two content platforms and merged to provide inter-modal mobility information. The content platform for individual transport integrates the existing data sources such as the police regional reporting office and also stores information about road works and major events. As one of the core functions the current traffic situation gets estimated by collecting current traffic volume and speed at about 180 urban and 200 motorway detection sites and utilizing this information in a strategic traffic model with roughly 8.000 links. It is a common approach to fit the result of traffic assignment to measurement values by adapting link impedance or travel demand.

In Berlin a different approach was chosen to estimate the current traffic situation by separating travel demand estimation and route choice in one step and the estimation of link volumes in a second step as the computation is done for different time horizons. Real-time route choice estimation is based on the Path Flow Estimator by Bell supported by an offline analysis of historical detector data to calibrate origin-destination matrices as a good starting solutions. Since the static assignment procedures embedded in the Path Flow Estimator is used for path estimation, the time horizon must not be significantly smaller than the maximum trip duration in the considered network. In time-steps of five minutes real-time link volumes are estimated by a propagation algorithm in which the estimated route choice is used to distribute detector values all over the road network. Link volumes and speeds are propagated separately. The propagation method relies on the fact, that the traffic volume observed at a detector is combined from the flows of a set of paths that spread onto the network before and after the detected link. If that path bundle for a measured volume is known, the portions of the single flows can be distributed in the road network along their paths. Since propagation becomes less accurate the more turning movements are incorporated, a reliability value is computed that decreases as distance from the measurement location increases.

Precise origin-destination tables are crucial to generate estimated link volumes and speeds within a given quality. Detector measurements of all 400 measuring sites were analyzed over a period of 4 months to find representative flow and speed profiles for all days of the week. In order to compare two days, the correlation was computed for the detector values of all 24 hours of the day separately. The correlated profiles were taken to cluster the days in typical days acting as representatives. Oflline matrix estimation had been applied for each of the representative flow and speed profiles to generate sets of hourly origin-destination matrices per representative. These OD-matrices were used as input for the Path Flow Estimator and in a second stage the propagation algorithms. The paper will present the propagation algorithm in detail.

A quality index has been computed by omitting single detectors while estimating link flow and speed and comparing estimated and counted values thereafter. The quality index considers Level-of-Service by using estimated volume and link capacity as input. The accuracy of estimated and observed LOS is within a 90%-range. Further results will be discussed within the paper. The accuracy of link speeds is important as these data is used within the dynamic route information system which is already under operation and improvement of accuracy is welcomed.

Publisher

Association for European Transport