Achieving Disaster-Recovery Scheduling in Real Time with a Supercomputer

How can recovery operations be planned quickly?

The situation after a large-scale disaster can change dramatically and rapidly, due to unforeseeable events such as secondary disasters and ruptured road networks. In such circumstances, quickly restoring lifelines and other infrastructures is crucial. As such, the timeline for the recovery work must be planned and put into action swiftly.

For this to be possible, it is necessary to determine the optimal plan in real time, taking into account live data on the disaster, geographical data and so on. However, such recovery plans have been difficult to calculate given the enormous amount of data, and the ever-changing nature of post-disaster conditions.

Devising Up-to-date Recovery Plans with a Local-search Algorithm

Kyushu University’s Institute of Mathematics for Industry (IMI) and Fujitsu Laboratories have now developed a system for scheduling disaster recovery in real time. This is a new technology that performs mathematical optimization on a supercomputer, efficiently formulating large-scale recovery plans that take into account complex, on-the-ground conditions. More specifically, it is a local-search algorithm that can efficiently choose the optimal work schedule from a vast number of possible schedules, while accounting for many complex constraints (task priorities, joint operations, priority of assigned areas, maximum working hours, etc.).

This contributes to optimized disaster recovery measures as it allows up-to-date recovery planning that reflects changing conditions, such as the spread of damage and recovery progress. Furthermore, this technology has potential applications in distribution and logistics, for scheduling deliveries and for allocating personnel accordingly. It is also possible to devise detailed delivery plans that take into account variable conditions, such as traffic congestion.

IMI and Fujitsu Laboratories are currently developing a platform for utilizing practical data that can gather real-time data on post-disaster conditions and recovery work. We are targeting this technology to be ready for practical use by local governments and emergency-response organizations from FY2017 onward.

Figure: Example of a recovery operation route (for 37 sites and 6 work teams)