Section oriented acquisition and application of microscopic traffic data using multi-sensor data fusion
Fazekas, Adrian; Oeser, Markus (Thesis advisor); Herty, Michael (Thesis advisor)
Aachen (2019, 2020)
Dissertation / PhD Thesis
Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2019
Road transportation and individual mobility are currently on the verge of fundamental technological change. The progress within this field is tightly bound with the advancements in sensor technology and the availability, quality and usability of the resulting road traffic data. Beyond aggregated traffic parameters like the flux, mean speed, and density used in macroscopic traffic analysis, future traffic applications will require a continuous location estimation of individual vehicles on a microscopic scale. Such data can be recorded with static radar detectors, laser scanners, or computer vision systems. In order to record the position data of individual vehicles over longer sections, the use of multiple sensors along the road with suitable synchronization and data fusion methods is necessary. To support decision-making in data driven application and to enable a systematic optimization of the data acquisition technique, a more specific analysis on the performance of vehicle position estimation has to be conducted. Furthermore, for a comprehensive discussion of the topic, the benefits of using microscopic traffic data need to be exemplified. Traffic safety analysis and traffic modeling are some of the critical research fields which can benefit of the availability of such data. This work addresses these issues by presenting new methods of recording, validating and applying highly detailed traffic data based on an efficient fusion of state of the art infrastructure based sensors. Within this thesis, appropriate methods for sensor fusion are presented, considering realistic scale and accuracy of the original data acquisition. As a first step, a closed formulation for a sensor offset estimation algorithm with simultaneous vehicle registration is presented. Based on this initial step, the datasets are fused to reconstruct microscopic traffic data using quintic Bezier curves. With the derived trajectories, the dependency of the results on the accuracy of the individual sensors is thoroughly investigated. The thesis also presents a systematic approach to measure the performance of position estimators based on a comparison of automatically recorded results with manually generated ground truth of individual vehicle positions in video segments. Furthermore, the methods described in this work enable a straightforward error handling, which is very useful in the development of position estimators. Based on the acquired and validated data, the derivation of a novel surrogate safety indicator is presented. The indicator is based on a Constant Initial Acceleration and reaction time assumption which considers the interaction between vehicles and describes the traffic safety of a road section. To examine the efficiency, the new developed indicator is compared to the original Deceleration Rate to Avoid a Crash (DRAC) and the modified indicator (MDRAC) which includes the reaction time. The results show that the new indicator is more sensitive in detecting critical situations than the other indicators and in addition describes the conflict situations more realistically. As an additional exemplification for the use of the acquired data, a novel two-dimensional first-order macroscopic traffic flow model is presented. The goal is to reproduce a detailed description of traffic dynamics for the real road geometry. In the presented approach, both the dynamics along the road and across the lanes is continuous. The result of the methods presented in this thesis constitute a comprehensive framework of acquisition, validation and use of microscopic traffic data. It includes a detailed discussion on the sensors, the resulting data and the current possibilities of recording high-detail traffic data. Hence, this thesis is a step towards the goal of developing an infrastructure able to support automized mobility in a safe, efficient and flexible way, while also enabling further research ideas that can be conducted based on this work.