UC San Diego

This dissertation develops a systematic model-based approach for the boundary control and estimation of freeway traffic congestion problem. Three topics of traffic congestion on a freeway segment are studied and include stop-and-go traffic oscillations, moving traffic shockwave, and downstream traffic bottleneck, which are governed by different partial differential equation (PDE) models and require the advancement and application of three PDE control techniques.

To supress stop-and-go oscillations, we introduce the macroscopic Aw-Rascle-Zhang traffic model, consisting of second-order nonlinear hyperbolic PDEs that govern dynamics of traffic density and velocity. The hetero-directional propagations of information in congested traffic generate the instabilities, motivating us to the stabilization problem for a coupled $2\times 2$ hyperbolic system. Using the backstepping method, a full-state feedback control is designed for ramp metering at outlet to actuate the outgoing traffic flow. We design boundary observer for state estimation and combine it with the full state feedback control to construct an output feedback controller. The observer design is validated with traffic field data. Under model parameter uncertainties, adaptive control design is proposed with on-line parameter estimation. Furthermore, we develop output feedback boundary control for two types of $4\times4$ nonlinear hyperbolic PDEs which arise from two-lane and two-class traffic congestion. Stabilization of two-lane traffic involves regulation of the lane-changing interactions with lane-specific varying speed limits while stabilization of two-class traffic tackles the heterogeneity of vehicles and drivers.

A moving traffic shockwave, caused by changes of local road situations, segregates light traffic upstream and heavy traffic downstream. This density discontinuity travels upstream. As a result, drivers caught in the shockwave experience transitions from free to congested traffic. The interface position is governed by an ordinary differential equation (ODE) dependent on the density of the PDE states, described with Lighthill-Whitham-Richards model. For the coupled PDE-ODE system, the predictor feedback design is applied to compensate the state-dependent input delays. We design bilateral boundary controllers to drive the moving shockwave front to a desirable setpoint position, hindering the upstream propagation of the traffic congestion.

Traffic on a freeway segment with capacity drop at outlet causes a downstream bottleneck. Traffic congestion forms because the traffic at the outlet overflows its capacity. Therefore the incoming flow of the segment needs to be regulated so that the outgoing traffic at the bottleneck area is discharged with its maximum flow rate. Since the traffic dynamics of the bottleneck is hard to model, we apply extremum seeking control, a model free approach for real-time optimization, to obtain the optimal input density at the inlet. The predictor feedback design is combined with the extremum seeking to compensate the delay effect of traffic state of the segment. The maximum flow rate is achieved at the bottleneck by regulating its upstream density at the inlet.