<![CDATA[Examination-document distribution is a time-critical institutional logistics problem, yet prior educational applications often examine fixed route instances and separate optimization from configurable decision-support workflows. This study develops a Python-based Capacitated Vehicle Routing Problem with Time Windows (CVRPTW) decision-support system that integrates directed OSRM matrices, Google OR-Tools RoutingModel, scenario parameters, and interactive route visualization. The solver applied PATH_CHEAPEST_ARC followed by GUIDED_LOCAL_SEARCH with a five-second search limit. Every examination site was mandatory, and directed road distance was the sole routing objective. Under the normal scenario, the best-found feasible solution used two vehicles and traveled 28.940 km. The practice-based reconstructed operational baseline, evaluated using the same matrix, traveled 34.686 km, corresponding to a model-based distance difference of 16.57%. When modeled as a single 100-package vehicle route under the study’s standardized assumptions, the reconstructed baseline was infeasible, with a 39-package capacity exceedance, modeled late arrivals at three campuses, and 107 minutes of modeled total lateness. Scenario-based estimated emissions were 6.76 kg for the reconstructed baseline and 5.64 kg for the CVRPTW solution using an assumed factor of 195 g CO₂/km. The tested deterministic scenarios remained feasible under the peak-demand condition of 176 packages, a 30-minute delay in depot-service start, and a traffic multiplier of 2.0. The contribution is a reproducible data-to-decision workflow rather than a new routing algorithm or proof of global optimality.]]>
