This project develops practical optimization methods for selecting, sequencing, and scheduling restoration actions for disrupted road networks based on incomplete and gradually improving information. Road networks may be severely damaged by events such as hurricanes and earthquakes, and prompt restoration is often necessary for the resumption of emergency services, other essential services, and normal activities.
The proposed methods employ artificial intelligence heuristics such as genetic algorithms and particle swarm algorithms to optimize the schedules of restoration tasks. A hybrid optimization approach combines fast traffic assignment with microscopic simulation to refine solutions. The methods are designed to start with incomplete, uncertain information and adapt dynamically as additional data becomes available from weather forecasts, work crews, and the public. The project also develops methods for pre-planning purposes, including preparing effective restoration plans based on estimated probabilities of disruptions and their consequences.
The research team will collaborate with the Maryland State Highway Administration and other agencies to ensure the practical applicability of the methods. Technology transfer activities include journal papers, conference presentations, software with a user manual, a final technical report, and workshops for interested transportation organizations.
Universities Involved
University of Maryland (Lead)
Virginia Tech
Principal Investigators
Paul Schonfeld
Hesham Rakha
Expected Research Outcomes & Impacts
The project will deliver optimization software and methods for road network restoration that address a significant gap in current practice. Currently used methods for prioritizing and scheduling alternatives in transportation networks are weak in considering interrelations among alternatives. The proposed methods evaluate entire sequences and schedules rather than individual alternatives, accounting for non-linearly additive benefits, costs, and constraints. The methods developed have broader applicability to comparable problems in manufacturing, logistics, communications systems, and emergency services.
Two Ph.D. students (one at UMD and one at VT) will receive advanced training in optimization and model development. The methods and results will be presented in classes and made available to interested agencies through workshops and a user manual.
Subject Areas
Network Optimization, Infrastructure Resilience, Disaster Response, Transportation Planning, Artificial Intelligence