A Model Based Approach to the Analysis of Intersection Conflicts and Collision Avoidance Systems

Intersection Safety;Left Turn;Driver Model;Speed Control;Mechanical Engineering;Engineering;Mechanical Engineering

Nobukawa, KazutoshiPerkins, Noel C. ;

University of Michigan

关键词: Intersection Safety;    Left Turn;    Driver Model;    Speed Control;    Mechanical Engineering;    Engineering;    Mechanical Engineering;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/89791/knobukaw_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

This dissertation studies the viability of driver assistance systems to improve the safety of „unprotected‟ left turns at signalized intersections. To achieve this, modeling and simulation have been conducted, including a driver model, with calibration and validation based on naturalistic driving data. A detailed analysis of the driving data has been conducted to reconstruct the vehicle trajectories in an automated manner. Particular challenges for this analysis include the development of automated detection of relevant events in a large database, automated estimation of sensor latencies, and the multiple application of Kalman filtering to fuse motion variables. A conflict analysis has been conducted to estimate the actual and predicted available gaps using the reconstructedvehicle trajectories. Monte Carlo simulations were conducted to create a large number of free left turn events in order to simulate a proposed driver assistance system and optimize safety performance. Optimization was conducted using multiobjective techniques which balance performance in terms of the rates of correct detections of conflicts, false alarms, and successful braking under the condition of correct detections based on Pareto optimality criteria.In this study, data to support the analysis was obtained from onboard instrumentation, where it was found essential to include detailed estimation of latencies between various sensors; after this, data fusion can be performed. It was found that high fidelity modeling of longitudinal control is critical to the safety system analysis. Also, it was found necessary to represent multiple levels of control, including visual preview and acceleration feedback. For the speed control reference, it was found that an ;;anticipated acceleration” can be used to define both straight braking events and free left turns; the driver may keep both options available during the intersection approach up to a critical decision point where the two references are equal. It was critical to the parametric optimization of the driver assistance system to take account of the need for warnings to be issued sufficiently early for the driver to respond; multiobjective design optimization was found to be an appropriate tool to include this requirement, as well as more typical requirements for involving false warnings.