Decision Optimization and Data-Driven Models for Agricultural Logistics in Resource-Constrained Environments: A Systematic Review of Approaches, Gaps, and Implementation Barriers in Sub-Saharan Africa

Agricultural logistics in Sub-Saharan Africa sits at a peculiar intersection: the models exist, the theory is well-developed, and the tools are increasingly available, yet postharvest losses for perishable crops remain between 20% and 40%, largely unchanged across two decades of research investment. This systematic review asks why. Using PRISMA 2020 methodology, 94 peer-reviewed studies published between 2015 and 2025 were screened and analysed across three categories: deterministic and metaheuristic optimization models, data-driven and machine learning approaches, and hybrid frameworks combining both. The results reveal a consistent pattern: optimization models are technically sound but built against assumptions that rarely hold in SSA, specifically reliable road networks, known harvest volumes, and stable demand. Machine learning applications are growing but largely confined to yield prediction and disease detection, with minimal deployment in logistics routing or real-time scheduling. Hybrid frameworks show the strongest theoretical potential but remain the least tested in African field conditions. Implementation barriers are not random; they cluster into four categories (infrastructural, technological, economic, and institutional) that interact in ways most individual models fail to account for. The review closes by proposing a staged deployment framework designed around the data and infrastructure realities of SSA rather than around the theoretical optima of the models.