UC San Diego
Traffic Throughput and Safety Enhancement for Vehicular Traffic Networks
- Author(s): Cho, Seokheon
- et al.
For decades, one of the most significant challenges in vehicular traffic networks has been how to mitigate traffic congestion characterized by increased vehicle queueing and slower traffic speeds. Whereas traffic congestion is a complicated and subjective matter, congestion delay consists of recurrent delay and non- recurrent delay. Recurring congestion is attributed to high excess demand that constitutes about half of congestion delay, and most of the rest is non-recurring congestion mainly occurring due to incidents. In order to reduce congestion more effectively in vehicular traffic networks, it is necessary to focus on how to control high utility demand for using traffic systems and how to guarantee safety against incidents by using various strategies, instead of inefficiently widening roads. Therefore, this dissertation provides diverse approaches to improve system throughput as well as to enhance collision safety for vehicular traffic networks through understanding of vehicular traffic flow, ramp-metering control, traffic safety metric, and inter-vehicle velocity control. In Chapter 2, we provide a macroscopic traffic flow model called a time-gap-based traffic model, which is a good and simple representative model for vehicular traffic flow. The proposed time-gap-based traffic model is well validated with empirical traffic data using least squares matching and is consistent with previous research outcomes about propagation velocity. Moreover, two analysis techniques to estimate the time gap from traffic measurement data are suggested. Chapter 3 presents two optimal coordinated ramp-metering algorithms based on the time-gap-based traffic model defined in Chapter 2. The purpose of the two optimization problems is to attain the maximum average system capacity for vehicular traffic networks. In addition, both of them are coordinated ramp- metering strategies that make use of the measurement data along a highway corridor to control all metered ramps simultaneously. One is a steady-state optimization problem used under supposition when the traffic system reaches a steady-state, whereas another optimal control is a time- variant optimization problem exploited when the traffic flow changes continuously with respect to time. The time- variant optimization problem considers an on-ramp queue management strategy. We propose a traffic safety metric called a safety marginal value (SMV) to be applied to continuous-space and discrete-time vehicular traffic networks in Chapter 4. The SMV represents the safety level of collision risk at every time step and is bounded by two non-negative integers. However, the computational complexity in determining the SMV grows dramatically, particularly when the number of vehicles traveling on a roadway increases. Hence, a finite space horizon for the SMV is also developed in order to prune the computational complexity of our proposed traffic safety indicator. A safety analysis of a car-following model is conducted with the SMV and with a microscopic traffic simulation. A car- following model called a target time-gap-based velocity update model is proposed in Chapter 5. Our proposed microscopic traffic flow model is used for inter-vehicle velocity control, which every vehicle exploits in order to refresh the velocity and position for the next time step. The microscopic traffic simulation results are matched well with empirical traffic data. Thus, the target time- gap-based velocity update model is considered a representative car-following model that can accurately mimic typical driving behavior. The effective domains of the target time gap and the update time interval, which guarantee both collision-free movement of all vehicles and system capacity enhancement compared to the traffic data measured in the field, are examined