SBIR Phase II: Improved Utilization through Artificial Intelligence and Prescriptive Logistics

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase II project is in improving distribution through rapid and reliable quality assessment and efficient logistics. To address this problem, the project is developing an integrated system that combines artificial intelligence for quality grading with prescriptive optimization to guide routing and utilization. This approach reduces rejected shipments and improves the recovery of surplus produce, increasing operational efficiency for both producers and commercial buyers. Beyond financial gains, the system will expand access to lower-cost food. Ultimately, this technology supports a more resilient and enduring food system by ensuring usable produce reaches appropriate consumers. The primary high-risk element of this project is the seamless integration of computer vision-based quality assessment with prescriptive analytics within highly variable produce supply chains. This challenge is significant because visual signals in produce are heterogeneous, and biological degradation introduces time-dependent uncertainty that is difficult to model. The technical innovation lies in the system's ability to not only interpret these complex visual states but to translate them into actionable prescriptive decisions under dynamic operational and logistical constraints. Successfully coupling real-time AI outputs with decision frameworks is a known "deep tech" hurdle that requires profound cross-domain integration. The project’s core intellectual contribution is the development of a unified representation of produce defects that generalizes across categories, paired with a decision framework that explicitly accounts for temporal decay and handling variability. This shifts the paradigm from static image classification to dynamic, decision-driven management of perishable assets. The methodology follows a rigorous, staged approach: first, curating diverse datasets across various produce types and seasonal conditions to capture real-world edge cases. AI models will be developed using transfer learning and structured data augmentation to ensure robustness and prevent overfitting. These quality signals will then be ingested by prescriptive optimization models that factor in storage conditions, transport constraints, and operational trade-offs. Performance will be validated through pilot deployments, measuring success via prediction accuracy, reduced shipment rejections, and optimized product utilization. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.