Project: #183 Dynamic Management of Food Redistribution for 412 Food Rescue Progress Report - Reporting Period Ending: March 30, 2019 Principal Investigator: Zachary Rubinstein Status: Active Start Date: July 1, 2018 End Date: June 30, 2019 Research Type: Applied Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2018 Mobility21 UTC Progress Report (Last Updated: March 30, 2019, 3:34 p.m.) % Project Completed to Date: 70 % Grant Award Expended: 70 % Match Expended & Document: 70 USDOT Requirements Accomplishments The focus of this period was to develop a system design for optimizing transport of donations, vetting it through 412's technical and user support personnel, and implement for piloting in the last quarter. Progress has been somewhat delayed, in part, due to 412 being a volunteer organization without dedicated development resources, but, overall, we believe it has been sufficient. We have developed a system model that has consensus with 412. The expected improvements are more targeted notification messages, i.e., reduced overall notifications, and more efficient use of volunteers based on travel distance and participation frequency. In this model, donations are allocated to volunteers based on location information provided by them. Respecting the privacy of the volunteers, there are different levels of location information they can provide. The best level in terms of scheduling donations is that the volunteer provide expected locations over the week. For example, imagine windows of availability annotated with locations for work and home. With this information, the scheduler can allocate based on where the volunteer is expected to be at the time of the transport. The second level is to provide current location by having the volunteer allow access to the localization services on her or his phone. This level has less accuracy than the first one because it says where the volunteer is now but does say where she or he will be at the time of the transport. The last level is to just use the home location of the volunteer, which is the most approximate with respect to where the volunteer will be at the time of the transport. Given a donation request, the scheduler returns a ranked list of volunteers that informs the order in which volunteers are notified that they have been selected for the donation. A volunteer must confirm before the donation is actually allocated to her or him, otherwise, the next group of volunteers (as decided by a policy) are notified. This model will greatly reduce the number of notifications volunteers over the current notification policy where, first, all volunteers within a set number of miles from the donation are notified, and then, if there are no takers, all volunteers are notified. The current status is that 412 and we have agreed upon a JSON-based API for the initial input of volunteer profiles, for donation requests, and for the returned ranked list of volunteers. 412's technology team is now working on supporting that API and adding the UI extensions needed for the elicitation of the new information. The anticipated completion date of these updates to the 412 infrastructure is the end of April, 2019. We are developing the scheduler supporting the new API and will be completed in advance of the expected integration with the 412 infrastructure at the beginning of May, 2019. 412 is also about to run a user survey with the model to get feedback from potential volunteers on the new model. We will incorporate any pertinent modifications of the system design based on this feedback. Impacts As the system has not been piloted, yet, the impacts are currently only speculative. As described in the Accomplishments section of this report, the expected major impacts will be a better user experience with the 412 app and a more efficient allocation of volunteers to donations. The new model will result in fewer notifications for volunteers, but we will need to assess the user experience and whether or not volunteer participation is affected. One of the challenges to evaluating the efficiency of the allocations is that, currently, the donations-to-volunteers ratio is heavily skewed towards volunteers, meaning that there are ample resources for transporting the donations. Efficiency becomes an issue when that ratio inverts, and volunteer become a constrained resource. The expectation is that 412's operations, both in Pittsburgh and in other cities, will grow in scale and that efficiency will become critical. One way to evaluate would be to do simulations, but that infrastructure does not currently exist and is beyond the scope of this project. In lieu of direct evaluation, we collect metrics on the resulting allocation and compare based primarily on models of the driving distances of the volunteers when servicing donations Other - API Document and System Design Outcomes New Partners None Issues None.