Funding source: US DOT

Contract number: 69A3551747109

Funding amount: $100,000
Performance period: 8/16/2020- 12/31/2021

Link to full seminar video: https://youtu.be/0tshE_qXdqg

The project objectives are:

The increasing penetration rate of GPS enabled personal mobile devices and connected vehiclesis leading to new sensing modes in urban traffic systems. This has spurred interest in methodologiesto estimate relevant traffic state such as traffic density, queue length, and travel time, to be usedas input to traffic control strategies, such as traffic signal control, variable speed limit, rampmetering, and routing. However, the impact of the correlation of measurements from these andtraditional sensing modes with relevant information about traffic state, and of the accuracy of real-time estimates on performance, e.g., in terms of energy consumption, emissions, throughputand travel time, is not understood rigorously. Correspondingly, there does not exist a principledapproach for joint choice of estimation and control algorithms to optimize performance of urbantraffic systems. In this project, we aim to develop methodologies to overcome this shortcoming.

We shall first provide analytical guarantees on key performance metrics for freeway and signal-ized networks, including travel time and emissions, under the commonly used approach to simplycombine independently designed estimation and control algorithms. We shall then consider si-multaneous choice of estimation and control algorithms that optimize end-to-end performance ofurban traffic systems, and which are amenable to distributed implementation. The key enablers ofour methodologies will be recent developments in feedback optimal and model predictive controltheory, and in adaptation of distributed optimization algorithms for control of traffic flow. Theanalysis will provide fundamental insights into how the quality of measurements, in terms of noiseand correlation with the underlying traffic state of interest, interplays with the traffic flow dynam-ics to determine performance of the overall data-driven control framework. Extensions to settingswhere key traffic parameters, such as wave speed, jam density and saturation flow, are unknownor uncertain will also be pursued. Case studies in PTV VISSIM for setups inspired by the LosAngeles area will augment the methodological contributions.