Scalable joint optimization of evacuation routes and schedules for large populations

Develop scalable algorithms that jointly optimize evacuation route assignments and departure schedules for city- or county-scale evacuation networks with large populations, producing high-quality solutions within reasonable computation time for objectives such as minimizing average evacuation time or evacuation completion time.

Background

The paper focuses on jointly optimizing evacuation routes and schedules under dynamic, confluent flow constraints, aiming to minimize system-level metrics such as average evacuation time and completion time. The authors note that while prior methods exist, even those providing bounded sub-optimality do not scale to city or county levels, making the problem computationally challenging at realistic scales.

To address scalability and congestion-dependent delays, the authors propose MIP-LNS (a heuristic that integrates mixed-integer programming with large neighborhood search) and MIP-LNS-SIM (which combines agent-based simulation with optimization to learn congestion-induced travel times). The explicit open problem statement motivates the need for such scalable methods capable of handling large populations within practical time limits.