Spatial Decision Support System for Evaluating and Optimizing Smart-Locker Networks

Parcel-locker networks are increasingly used to address last-mile delivery inefficiencies, yet their performance depends strongly on whether locker locations provide accessible and spatially balanced service. Although previous studies have examined parcel-locker siting through facility-location models, multicriteria decision-making, simulation, or accessibility analysis, fewer studies provide an integrated decision-support workflow that evaluates an existing real-world network, statistically diagnoses spatial underservice, and uses these diagnostics to guide scalable expansion. This paper addresses this gap by developing a cloud-based spatial decision support system (SDSS) that combines spatial-statistical evaluation with genetic algorithm-based (GA) p -median optimization. The framework first uses descriptive accessibility metrics and kernel density estimation to characterize current siting patterns, then applies local Moran’s I (LMI) to identify statistically significant clusters of high population-weighted access burden. These underserved clusters are used both as diagnostic outputs and as a reduced candidate pool for optimization. The method is tested on a real smart-locker network in the Athens Metropolitan Area, comprising 858 locker locations and 36,268 population parcels. Results show that adding 40 lockers, approximately a 5% network expansion, increases population coverage within a 300 m service radius by about 3%. The proposed LMI + GA workflow reduces the candidate search space from 36,268 parcels to 5,346 underserved parcels and reduces runtime from about 2 min to under 20 s while maintaining comparable solution quality. The SDSS provides an accessible, data-driven environment for identifying coverage gaps, prioritizing expansion sites, and supporting strategic smart-locker planning.