Optimising Daily Waste Collection Routes: A Case Study of a Dutch Waste Collection Operator

Dutch waste collection operators currently plan their daily routes manually or with rudimentary tools, leaving open the question of how much operational efficiency could be gained by applying formal vehicle routing optimisation. This thesis addresses that question through a case study of a Dutch waste collection operator, using six months of disposal records covering the first quarters of 2024 and 2025. The collection problem is modelled as a Multi-Depot Capacitated Vehicle Routing Problem with Intermediate Facilities (MDCVRP-IF), incorporating driver-break requirements and shift-time windows, and solved using PyVRP, a hybrid genetic search metaheuristic. A pilot run shows distance savings between 11% and 32% relative to observed practice, depending on the assumed vehicle capacity and the resulting fleet size. The thesis then extends the model to operational conditions, where customer demands must be estimated in advance rather than observed ex post. Several estimators for waste demand are evaluated on a rolling train/test split, with overflow events resolved by a recourse rule and the resulting realised distance compared against both observed practice and the perfect-information optimum. The thesis quantifies how much of the theoretical saving survives the transition from perfect information to demand uncertainty, and identifies which estimator design choices matter most in practice