Comparison of waste bin placement optimization methods: A case study in Zurich

Recycling participation declines sharply when collection points are inconvenient to reach. In Zurich, Switzerland, over 29,000 residents (6.5 % of the population) live beyond the 10-minute walking threshold set by the municipal waste authority. This study developed an accessibility-based siting framework for identifying the minimum number and best locations of additional Recycling Collection Points (RCPs) within the 10-minute target for Zurich residents. Three methods were compared: two Density-Based Spatial Clustering of Applications with Noise (DBSCAN) approaches and a Mixed Integer Linear Programming (MILP) p-median formulation. All use elevation-aware network routing rather than Euclidean approximations, and all select sites from the same screened candidate set of 305 car park polygons; results are therefore optimal conditional on this candidate set, not over Zurich's full urban surface. MILP optimization, minimizing population-weighted walking duration, outperformed both clustering variants by 2-5 % on coverage metrics. Adding 12 sites reduces the underserved population by approximately 30 % (from 29,320 to 20,451 residents); the city-wide average walking time changes only marginally, from 4:56 to 4:36, because the gain is concentrated in the underserved tail rather than spread across already well-served residents. Clustering, while less effective for site selection, is useful for diagnosing spatial gaps. The framework optimizes pedestrian accessibility only; collection-vehicle routing, container capacity, and operational costs are out of scope. Future work will extend the framework to multi-objective optimization that integrates these factors and to agent-based demand projections for longer planning horizons.