<p>Inefficient waste collection and transportation systems in Morocco have created significant operational and environmental challenges, highlighting the need for advanced optimization techniques to reduce resource consumption and carbon emissions. This study models the waste collection process as a Vehicle Routing Problem with Time Windows (VRPTW) and proposes a novel two-phase optimization algorithm, the Clustered Modified Large Neighborhood Search (CMLNS). Unlike conventional methods, CMLNS combines clustering-based initialization with an enhanced neighborhood search strategy. In the first phase, a k-means heuristic clusters locations based on their temporal attributes, followed by routing the clusters using four insertion operators. The second phase applies the Modified Large Neighborhood Search (MLNS), which randomly employs three removal operators and selects the best of four insertion operators in each iteration. The MLNS also integrates a blur parameter that improves exploration and the hill-climbing mechanism that improves exploitation. Extensive experiments on the classic Solomon benchmark instances demonstrate that CMLNS achieves competitive performance compared to state-of-the-art algorithms by reaching a 0.9% gap with the Best Known Solutions (BKS), confirming its robustness and efficiency. These findings suggest that the CMLNS algorithm offers a promising approach to optimizing Morocco’s waste management systems, enhancing both operational efficiency and environmental sustainability.</p>