Online traffic state estimation based on floating car data

📝 Abstract

Besides the traditional data collection by stationary detectors, recent advances in wireless and sensor technologies have promoted new potentials for a vehicle-based data collection and local dissemination of information. By means of microscopic traffic simulations we study the problem of online estimation of the current traffic situation based on floating car data. Our focus is on the estimation on the up- and downstream jam fronts determining the extension of traffic congestion. We study the impact of delayed information transmission by short-range communication via wireless LAN in contrast to instantaneous information transmission to the roadside units by means of mobile radio. The delayed information transmission leads to systematic estimation errors which cannot be compensated for by a higher percentage of probe vehicles. Additional flow measurements from stationary detectors allow for a model-based prediction which is effective for much lower floating car percentages than 1%.

💡 Analysis

Besides the traditional data collection by stationary detectors, recent advances in wireless and sensor technologies have promoted new potentials for a vehicle-based data collection and local dissemination of information. By means of microscopic traffic simulations we study the problem of online estimation of the current traffic situation based on floating car data. Our focus is on the estimation on the up- and downstream jam fronts determining the extension of traffic congestion. We study the impact of delayed information transmission by short-range communication via wireless LAN in contrast to instantaneous information transmission to the roadside units by means of mobile radio. The delayed information transmission leads to systematic estimation errors which cannot be compensated for by a higher percentage of probe vehicles. Additional flow measurements from stationary detectors allow for a model-based prediction which is effective for much lower floating car percentages than 1%.

📄 Content

arXiv:1012.4567v1 [cs.OH] 21 Dec 2010 Online traffic state estimation based on floating car data Arne Kesting1 and Martin Treiber1 Abstract B esides the traditional data collection by stationary detectors, recent advances in wireless and sensor technologies have promoted new potentials for a vehicle- based data collection and local dissemination of information. By means of mi- croscopic traffic simulations we study the problem of online estimation of the current traffic situation based on floating car data. Our focus is on the estima- tion on the up- and downstream jam fronts determining the extension of traffic congestion. We study the impact of delayed information transmission by short- range communication via wireless LAN in contrast to instantaneous information transmission to the roadside units by means of mobile radio. The delayed in- formation transmission leads to systematic estimation errors which cannot be compensated for by a higher percentage of probe vehicles. Additional flow mea- surements from stationary detectors allow for a model-based prediction which is effective for much lower floating car percentages than 1 %. 1 Introduction Detailed and reliable online traffic state estimation is a prerequisite for advanced traffic information systems and the next generation of driver assistance sys- tems [1, 2]. Recent advances in wireless and sensor technologies have therefore promoted new potentials for a vehicle-infrastructure integration (VII) system which allows for a wireless communication between roadside sensors and vehicles equipped with communication interfaces [3, 4]. Data gathered from equipped ve- hicles (‘probe vehicles’ or ‘floating cars’) can be used to assess and predict traffic conditions and, in turn, information about incidents, travel times or congestion can be communicated from roadside units to the vehicles [5, 6, 7, 8]. Technische Universit¨at Dresden, Institute for Transport & Economics, W¨urzburger Str. 35, 01062 Dresden, Germany 1 2 A. Kesting and M. Treiber In this paper, we consider the following setup for an online traffic-surveillance application based on floating car data: Probe vehicles collect their positions and speeds over time and communicate this information to roadside units either periodically by instantaneous forwarding by mobile phone communication, or, alternatively, by short-range communication when passing the location of a road- side unit. In any case, the roadside units are connected to each other and collect the data periodically for an online traffic state estimation. More specifically, we consider the prediction of the upstream jam front which, for example, can be used to warn the driver when congestion is ahead. For this purpose, we study the data collection, the communication to roadside units, and the traffic state estimation ‘in-the-loop’ by means of microscopic traffic simulations as a function of the percentage of equipped vehicles. This approach enables us to compare the estimates of both communication modes to the – known and reproducible – traf- fic situation. Since the expected equipment levels in the first deployment phase will be small, we consider small percentages of probe vehicles below 3 % only. The paper is structured as follows: In Sec. 2, the microscopic simulation set- up is described. In Sec. 3, the algorithm for the estimation of jam fronts is presented and a measure for evaluating the estimation quality is defined. In Sec. 4 an alternative model-based approach is presented which uses additionally flow measurements from stationary detector. We close with a discussion in Sec. 5. 2 Microscopic Simulation Setup 2.1 Reference Scenario with Empirical Boundary Conditions For the following case study of online estimation of jam fronts one needs a suited and realistic traffic situation serving as reference scenario. For this purpose we start with the empirical traffic jam observed on the German freeway A5 shown in Fig. 1(left). In the simulation environment, we consider a road-section with three lanes and a flow-conserving bottleneck [9]. The empirical traffic flows and truck percentages from one detector location (shown in Fig. 2) serve as upstream boundary conditions assuring realistic traffic volumes and degrees of traffic het- erogeneity. The simulator uses the Intelligent Driver Model [9] as a simple, yet realistic, car-following model, and the general-purpose lane-changing algorithm MOBIL [10]. The parameters of the car-following model (see Table 1) and the bottleneck strength (modeled by a local increase of the time gap parameter T by 40 % in a given time interval) have been adapted in order to reproduce the empirically observed spatiotemporal congestion pattern on a semi-quantitative level. Figure 1(right) shows the simulation result. 2 Microscopic Simulation Setup 3 A5 North, April 11, 2001 8 8.5 9 9.5 10 10.5 11 t [h] 480 482 484 486 488 490 x [km] 0 20 40 60 80 100 120 140 V[km/h] Fig. 1 Empirical traffic state from the German freeway A5 reconstructed

This content is AI-processed based on ArXiv data.