University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies
Henry Liu

Henry Liu

Model Reference Adaptive Control Framework for Real Time Traffic Management Under Emergency Evacuation

Presentation by Henry Liu, Assistant Professor, Dept. of Civil Engineering

September 27, 2005

Evacuating metropolitan areas can be a traffic operations nightmare, as evidenced by the massive traffic jams during this summer's evacuations of hurricane-menaced New Orleans and Houston. Civil engineering professor Henry Liu outlined his approach to improving traffic management for evacuations at the Sept. 27, 2005 Advanced Transportation Technologies Seminar, sponsored by the ITS Institute.

Liu asserted that evacuation efforts often run into problems despite elaborate scenario-based planning efforts carried out in advance, because planning can't predict all possible scenarios. Instead of scenario planning, Liu's work focuses on real-time traffic management, or, as he put it, "how can we respond to this specific disaster, now."

But by their very nature as extreme events, emergency evacuations resist traditional network evaluation methods, because standard assumptions about network use patterns do not apply. Traffic during evacuations is dominated by movement to "safe zones" and fraught with unpredictable behavior by frustrated and panicked motorists. However, emergency situations do provide the authority for strict centralized control of the traffic system, using signals and traffic control officers on the ground to direct vehicles as needed.

These characteristics led Liu to develop an adaptive control approach to evacuation traffic management. Adaptive traffic control systems, which use current traffic data to change their control behavior in response to changing conditions, offer the ability to compensate for disruptions by rerouting traffic along optimal routes. A reference model of the road network and a traffic simulation engine calculate the overall system objective using current data.

Liu explained how he simulated a stripped-down version of this system (without origin-destination estimation and resource allocation modules) on a simplified road network containing only a few nodes and links. For each node (intersection) in the network, the system calculates the optimal turning percentage to achieve a desired performance measure, such as lowest total travel time or minimum number of "victim vehicles" (vehicles unable to clear the network before a deadline). The system uses a "rolling horizon" approach to estimate traffic flow characteristics every two minutes and update its control strategy accordingly.

A complete system of this type, Liu said, would comprise five major components: the reference model; a simulation engine capable of computing the routes of individual vehicles; the traffic signal controller; an origin-destination estimation module; and a resource allocation module (to optimize performance in the event that not all areas can be effectively controlled).

Liu, who joined the Department of Civil Engineering as an assistant professor in August of 2005, began working with adaptive traffic control systems before coming to Minnesota; he said he intends to continue his work here at the University of Minnesota.