Multi-objective location-routing optimization based on machine learning for green municipal waste management

Most of the existing municipal waste management (MWM) systems focus on the optimization of the waste disposal center locations and waste collection paths, which can be modeled based on the location-routing problem (LRP). This study models a green MWM system by a three-objective location-routing problem to achieve equilibrium among the total cost, carbon emission, and residential satisfaction. The amount of waste demand for each customer is considered as an independent discrete random variable following a normal distribution. The multi-objectives and non-deterministic characteristics make this problem more intractable than the traditional LRP. A multi-objective optimization algorithm based on decision tree classifier is proposed for solving this problem. The decision tree classifier learns from previous searching experience, and then guides the following evolution process to avoid blind searching. The experimental results show that the proposed algorithm has high competitiveness compared with other state-of-art methods. A case study is also conducted for a real waste collection problem in a certain area of Beijing. The proposed method adopts efficient location-routing strategies to balance the total cost, carbon emissions, and distance between residential areas and waste disposal centers.