Design, Field Implementation and Evaluation of Adaptive Ramp Metering Algorithms: Final Report

Part I: Modeling, Estimation, and RampMetering Control of Freeway Traffic Themain objective of Part I is to develop improved algorithms for freeway on-ramp metering control. Toward this end, two macroscopic traffic models are first introduced and analyzed. The first is a modification of the cell transmission model (CTM) developed by Daganzo. This modified CTM uses cell vehicle densities as its state variable and allows nonuniform cell lengths. The second is a piecewise linearization of the modified CTM, which results in a switching-modemodel (SMM). The linear structure of the SMM simplifies system analysis, data estimation, and control design. Furthermore, different observability and controllability properties in different modes motivate the design of a switching ramp-metering controller. A mixture Kalman filter based freeway state estimator is designed. This estimator is able to provide estimated vehicle densities at unmeasured locations, as well as the most probable traffic congestion modes (free-flow or congested), which are not directly observed. The test results on a short freeway section show that, on average, a mean percentage error of about 10% in density estimation is achieved, and the performance is consistent over traffic data collected on different days. This estimator is then applied to the Interstate 210 West (I-210W) test site in Pasadena, California. A suite of ramp-metering control algorithms are developed, including 1) a switching ramp-metering controller, which adapts to different traffic dynamics under different congestion conditions and is designed using a multirate LQI (Linear Quadratic control with Integral action) approach, 2) a queue-length estimator and regulator that estimates on-ramp queue lengths using the available queue-detector speed data and achieves a better performance over the currently used “queue-override” scheme, and 3) a local control strategy to achieve the goal of reducing the spatial and temporal extent of the congestion. Simulation results on macroscopic CTM and microscopic VISSIM models that are calibrated to the I-210W test site demonstrate the performance and effectiveness of the algorithms. In the CTM simulations, the switching ramp-metering controller accurately tracks the mainline density profiles. In the VISSIMmicroscopic simulations, the Total Vehicle and Passenger Delays are both reduced by 16% when queue length constraints are strictly enforced and by 20% when queue lengths length constraints are not enforced. In this simulation study, the switching ramp-metering control outperforms both ALINEA, a well-known local traffic-responsive ramp-metering algorithm introduced by Papageorgio et al., and SWARM, a combination of coordinated and local traffic-responsive ramp-metering algorithms developed byNETCorporation, when they are tested under similar configurations.