With the popularity of waste classification and increasing requirements for coordinated waste management towards sustainability, it is essential to study how to generate efficient waste collection solutions for classified treatment towards social-economic and environmental goals. We proposed a new methodological framework to identify not only the optimal routes for different collection modes in classification but also the most desirable mode under different decision preferences, by a novel multi-objective optimization model applicable to different collection modes with addressing economic, environmental and social dimensions of sustainability. We incorporated the lifecycle carbon emissions of fuel and the economic and environmental impacts of vehicle idling due to queening to make the model more realistic and comprehensive. To solve the model, we proposed a modified metaheuristic algorithm based on particle swarm optimization and genetic algorithm. Through the case study in a pilot waste classification city of China, we demonstrated the effectiveness of the designed algorithm and the applicability of the proposed methodological framework. The results indicate that there are strong synergies between environmental and economic objectives, but trade-offs between social and environmental/economic objectives. “Classification then combined-collection” should be given priority under economic and environmental preferences, while “Classification then separated-collection” should be adopted under social preference.