Traffic parameter estimation on motorway networks by combination of filtering techniques Yubin Wang Faculty of Technology, Policy and Management Delft University of Technology Delft, The Netherlands e-mail: yubin.wang@tudelft.nl Yufei Yuan Faculty of Technology, Policy and Management Delft University of Technology Delft, The Netherlands e-mail: y.yuan@tudelft.nl Jos Vrancken Faculty of Technology, Policy and Management Delft University of Technology Delft, The Netherlands e-mail: j.l.m.vrancken@tudelft.nl Abstract—In order to perform road traffic control, it is very important to estimate the traffic parameters which can not be measured directly from sensors. In this paper, we will focus on turn fraction estimation based on a new road network representation which is used in traffic control software at the Dutch traffic management company Trinit´ e Automatisering B.V. The common approach for the turn fraction estimation is by applying Kalman filter. However, the sensor information for motorways is not always available due to the fact that there are no physical sensors or detector failure on some parts of motorways. In this case, Kalman filter can not be applied to estimate turn fraction. A new approach by combining of Kalman filter and a low pass data filtering technology called Treiber-Helbing-filter is presented. This approach can contribute solving the problem by using Treiber-Helbing-filter to complete the missing data firstly. Then, turn fraction is able to estimate by using Kalman filter and visualize in traffic control software. Index Terms—Traffic Control, Traffic Parameters Estimation, Turn Fraction Estimation I. I NTRODUCTION Traffic jams are a big problem for many countries. They cause fuel wasting, pollution, time losses, have a great impact on the economy, the environment and the quality of life. One possible solution to reduce traffic problems is to build new infrastructure. However, there are many difficulties with this approach such as high cost and lack of space. Traffic control is an attractive alternative solution to reduce traffic problems due to its relatively low cost. Road Traffic Management is being applied in many countries for well-known purposes such as congestion prevention, augmenting efficiency by minimizing travel times, improving traffic security and driving comfort, and reducing environmental damage. Road Traffic Control (RTC) is one of the main activities within road traffic management, next to demand management, incident handling and pricing. RTC is about influencing traffic streams in order to improve traffic flow. Traffic data collection from road side equipments is essential for RTC. Traffic parameters (like turn fraction and OD matrix, ect.) which can not be measured directly are also very important for RTC. RTC can be improved if as much (accurate) as possible traffic parameters can be estimated. In this paper, we will focus on traffic turn fraction estimation on motorway networks. The literature offers an abundance of approaches to estimate turn fraction, such as Likelihood methods [5], Bayesian estimator [10] and Kalman filter [1]. Most common approach is based on Kalman filter. We mention only a few. Simulation of turn fraction based on the unconstrained and constrained Moving Horizon Estimation (MHE) Kalman filtering is presented in [3]. Traffic state estimation based on extended Kalman filter is presented in [9]. Most studies are simulation based. However, in practice the necessary traffic information from road side equipment is not always available for the purpose of turn fraction estimation. In this paper, we will present the approach combining Kalman filter and a data filtering technology called Treiber- Helbing-filter [6] to contribute solving the above problem for motorway. First, Treiber-Helbing-filter is used to complete the missing data on motorway. Then, Kalman filter is used to estimate turn fraction on motorway. By applying this approach to traffic control systems, turn fraction is able to estimate. This approach is currently being implemented in a road traffic network management software which based on a new road network representation. The system was developed by the Dutch traffic management system company Trinit´ e Automatisering B.V. Most of the works have been made by applying new algorithms in a simulation environment. Our approach is challenging in the sense that the algorithms were implemented in a real system and try to solve traffic problems in practice. Proceedings of the 2009 IEEE International Conference on Systems, Man, and Cybernetics San Antonio, TX, USA - October 2009 978-1-4244-2794-9/09/$25.00 ©2009 IEEE 3658