Engineering researchers at North Carolina State University have developed a mathematical framework that can be used to help hunger-relief organizations get food to households that need it more efficiently than conventional methods. The advance, which has already been incorporated into an app, could also lead to improved efficiency for other businesses that face logistical challenges associated with deliveries and volunteer assignments.
"Food banks serve a critical function, helping households in their communities address hunger," says Leila Hajibabai, corresponding author of a paper on the work and an associate professor in NC State's Edward P. Fitts Department of Industrial and Systems Engineering. "But delivering food to those households poses significant logistical challenges.
"Food availability varies from day to day, as do the households that require support. And there are a host of variables that can change even over the course of a day. Requests for food can come in at any time; households may only be able to receive deliveries at certain times of day; delivery drivers might not show up for a shift or be unable to complete all of their assigned deliveries."
"Our goal was to create an optimization framework, which is a mathematical tool that can be used to determine the most efficient way to assign deliveries to drivers and establish the route that will allow them to complete those deliveries as efficiently as possible," says Mehr Salami, first author of the paper and a Ph.D. student at NC State. "In addition, we wanted to make sure the tool is dynamic. We need it to both respond to changing variables over the course of the day and predict possible needs that might arise in order to maximize efficiency."
For this project, the researchers worked with a regional food bank to collect data on its delivery network: number of drivers, vehicles, food availability, household demand and so on. The researchers then used that data to solve the problem in a way that accounts for all of the relevant variables.
"At the beginning of a day, our framework identifies an efficient way of distributing deliveries among the available drivers and the routes they should take," Salami says. "But it doesn't stop there. Instead, the framework takes into account updates that come in over the course of the day. In addition, the framework runs a series of predictions that try to forecast household requests over the course of the day in order to make informed decisions about when a driver will need to add deliveries to their route or return to the warehouse to pick up more food."
To see how well the framework performs, the researchers compared its performance against three established benchmark methods.
The new optimization framework did a better job than the benchmarks in creating more efficient distribution of tasks among drivers, getting food to households more efficiently, and accounting for unexpected requests from households. The benchmark methods were more computationally efficient; however, the proposed framework still produced results in less than a minute while serving more households.
"From a practical standpoint, our framework was more useful for the food bank," says Salami.
"We have also developed an app that incorporates the optimization framework which we are hoping to make freely available to nonprofit food banks in the near future," says Hirumi Niwunhella, a Ph.D. student at NC State who is not a co-author on the paper but who played a critical role in creating the app.
"We are also weighing the creation of an app that could be used by businesses to address other delivery logistics challenges and are in the process of preserving our intellectual property for that application," says Hajibabai.
The paper, "Anticipatory Monte Carlo Tree Search-Based Optimization for Stochastic Dynamic Routing with Time Windows," is published open access in the journal Computer-Aided Civil and Infrastructure Engineering. The paper was co-authored by Kuangying Li, a Ph.D. graduate of NC State who is now an assistant professor at Wuhan University of Technology.
This work was done with support from the National Science Foundation under grant 2125600.
