Online Large-scale Garbage Collection Scheduling: A Divide-and-conquer Approach

Online garbage collection scheduling is demanding for large cities to reduce the increasing operational costs. However, the garbage collection problem is NP-complete, making the problem intractable when the number of garbage sites is large. In this paper, we first intensively investigate the garbage collection problem and derive insightful theoretical guidance for decomposing a large-scale garbage collection problem. We then propose an agglomerative hierarchical clustering algorithm, called Pie, for online large-scale garbage collection scheduling, where the original problem can be equivalently decomposed into a set of small-scale tractable sub-problems. We implement Pie which has a O(n2) complexity and adopt LKH-3, the state-ofthe-art CVRP algorithm, as the underlying algorithm to solve sub-problems obtained by Pie. We conduct extensive trace-driven simulations on 11 real-world datasets. The results show that Pie can effectively reduce both the overall collection cost and the running time, demonstrating the efficacy of the Pie algorithm.