Project: #183 Dynamic Management of Food Redistribution for 412 Food Rescue Progress Report - Reporting Period Ending: Sept. 30, 2018 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: Sept. 30, 2018, 9 a.m.) % Project Completed to Date: 25 % Grant Award Expended: 25 % Match Expended & Document: 25 USDOT Requirements Accomplishments In this first quarter, the focus was on familiarizing our team with current 412 Food Rescue operations and to develop an initial approach to optimizing those activities that would have the most impact and would be readily adopted into their current work structure. To that end, we have completed the following steps: - Worked with Leah Lizarondo, Co-Founder and CEO of 412 Food Rescue, and 412 staff to familiarize ourselves with their current technology infrastructure and mobile application for crowd sourcing the transportation of food donations to non-profit distributors. Held a Kickoff meeting for the project and followed up with additional discussions. - Learned about 412's current dispatching process, where they are informed of a donation and then determine both a destination for the donation and transportation resources. As part of this on-boarding process, we became a dispatcher and made real-time decisions for actual donations. - Based on 412's current decision-making process, we determined to treat the donation-destination problem and the transportation problem separately, where the transportation problem already has a destination specified in the request. A better solution might result if both problems were considered as one, but 412's operational structure would have to change substantially to consider that more complex problem. Initially, we will focus on the transportation problem. - Worked with 412 staff to identify priorities in decision-making process, both for how donations are assigned to non-profit distributors and how transporters are assigned to donation-movement requests. More discussion of this refinement is provided below in impacts. - Met with Technology Board for 412 and presented the project. The Technology Board comprises a set of advisors and leaders of companies that they outsource their current technology requirements. We agreed to a high-level interface where our technology can integrate with the current platform. We are currently working on a more detailed document. - Met with Min Kyung Lee and Ariel Procaccia, who have a Uptake grant for developing a social-welfare-based system for suggesting destinations for donations, to determine synergies between our projects. We decided that their work can be used to inform destinations for transportation requests, which will be serviced by our project. Impacts As it is relatively early in the project, we have no social impacts but there has been one of academic interest that has also affected some of the decision-making made by 412. Before our joint discussions, we considered the objective function, i.e., the goal of the optimization, to be to allocate resources as efficiently as possible based on their availability and minimizing the distance traveled. But, after discussion with 412, two other factors became apparent. First, there is a priority to limit the number of notifications a transportation volunteer receives about new requests. A second priority is ensuring that people get assigned to requests with some regularity to keep them engaged in the process. Both of these priorities can work cross purpose with minimizing distance traveled, so there needs to be some kind of formal determination of when to make that tradeoff. 412's current operation only coarsely addresses these requirements. They use a two-phase allocation procedure based on the the current position of a volunteer relative to the donation pick-up location. The first phase is to determine a group of volunteers to notify first, i.e., those that are within some delta of the pick-up, and send them the notification of a request. Note that a notification is sent without respect to a temporal dimension, e.g., without considering that the pick-up is later in the day and that the current position of the volunteer may not be relevant. Also, the destination is not considered at all. If no assignment can be made after the first phase, then, in a second phase, the notification is sent to all volunteers in the network. While this approach does do some management of the notifications being sent and has some consideration of spatial proximity, it is fairly approximate. We are currently working on a detailed system design that should result in more efficient schedules while better meeting these other two requirements. The high-level architecture will allow volunteers to optionally specify their availability, that is, periods of time that they are available to transport and their expected locations. For example, a person might specify times they are home and times they are at work. These availabilities will allow for more targeted volunteer assignments based on both spatial and temporal characteristics. In addition, volunteers can specify whether or not they are willing to let the app automatically negotiate for assignments without notifying the person unless they are selected for the assignment, in which case the person can decide whether or not to accept. This feature should result in fewer notifications for volunteers. There is still an open question about how to address the equity requirement to encourage participation. There would be two classes of users, those that opt in to automatic negotiation and those that do not. If the assignment process always checks the automated ones first, then the non-automated ones could be starved. One possible approach is to periodically to have notifications sent out more broadly, but this issue is still under review. Other None as of this report. Outcomes New Partners No new partners Issues No major changes.