November 8, 2017
Modeling, Control and Estimation of Traffic Networks
Abstract:
This talk discusses some of our recent advancements in control and estimation of road networks. Traffic congestion is a major source in efficiency in both the developed and developing worlds, with one study estimating that in 2014, delays due to congestion cost 7 billion hours and $160B in the US alone. However, mitigating congestion through control techniques is difficult, as traffic congestion exists in a confluence of complex phenomena, such as nonlinear shockwaves, emergent macroscopic network effects from multiple agents, and low system observability and controllability. Growth of traffic demand shows no sign of decreasing, so continued infrastructure expansion must be combined with continued development of traffic control engineering to abate these societal costs. Some of today's traffic control efforts make use of novel formulations of these nonlinear systems and new sources of data provided by the connected and autonomous vehicles now entering the fleet.
In the first part of talk I will describe a set of modeling and simulation tools for traffic operations planning to provide quick and quantitative assessments of the benefits that transportation management center control policies can provide on freeway corridors, in order to decrease congestion. These tools are based on a self-calibrated Asymmetric Cell Transmission Model (ACTM) traffic macroscopic simulator, which is based on a well-accepted theoretical model of traffic flow; it is parsimonious and does not require parameters that cannot be estimated from traffic data; and has been tested for reliability on several freeways. In addition to describing some basic controllability and observability properties of this model, I will also briefly describe a set of parameter calibration, ramp flow estimation and sensor fault detection algorithms that were developed in order to achieve reliable simulation of freeway traffic.
In the second part of the talk, I will focus on traffic control, in order to ameliorate congestion. I will first show how the problem of minimizing a cost functional composed by the linear combination of total travel time and total travel distance in a freeway described by the ACTM can be solved as a linear program. Subsequently, I will focus on implementations of freeway ramp metering using Model Predictive Control (MPC) algorithms and will address issues related to guaranteeing the persistent feasibility and stability of the algorithms. Thirdly, I will present a novel framework for freeway ramp metering that is based on maximizing the aggregate utility of onramp flows. We show how solving the dual problem of maximizing the network utility via a gradient projection algorithm synthesizes a low- complexity control law that is simple enough to be implemented on real platforms, while being robust to measurement noises. Our control algorithm is distributed as at each time step, every onramp selects a traffic flow to maximize its own benefit, and the network adjusts unit traffic flow prices for different onramps. Finally, I will extend the network utility  design framework for the simultaneous signalization and joint perimeter control of arterial traffic networks.
Biography
Roberto Horowitz is the current chair of the Department of Mechanical Engineering at UC Berkeley and holds the James Fife Endowed Chair in the College of Engineering. He received a B.S. degree with highest honors in 1978 and a Ph.D. degree in 1983 in mechanical engineering from the University of California at Berkeley and became a faculty member of the Mechanical Engineering Department in 1982. Dr. Horowitz teaches and conducts research in the areas of adaptive, learning, nonlinear and optimal control, with applications to Micro-Electromechanical Systems (MEMS), computer disk file systems, robotics, mechatronics and Intelligent Vehicle and Highway Systems (IVHS). He is a former co-director of the Partners for Advanced Transportation Technology (PATH) research center at U.C. Berkeley. Dr. Horowitz is a member of IEEE and ASME and the recipient of the 2010 ASME Dynamic Systems and Control Division (DSCD) Henry M. Paynter Outstanding Investigator Award.