Time Window and Location Based Clustered Routing with Big and Distributed Data

In this paper, a novel vehicle routing algorithm will be presented. Proposed method will be based on “time windows-based clustering” and “location-based clustering”, applied in reversable consecutive order. The method partitions and models the solution space with machine learning technologies, resulting in a better performance for time window and geospatial clustering calculations. Routing process, on the other hand, will be built upon already present open source tools, giving it usability, applicability, manageability, and integration perspectives. The process combines “cluster+cluster+route” units with post process enhancements. Previous works on location-based clustering are proved to be successful, albeit with some disadvantages. On the other hand, routing algorithms have mostly implemented time window calculations as second-class citizens. In this method, time window is a major ingredient of the modelling process. This paper will also differs from some other combinatoric methods used in literature. A history and general description of used methods and tools will also be provided. It is shown that the algorithm can generate good results, some of which are the best values in the recorded literature so far. The method is applied on a big data platform. Horizontal scaling and distributed processing capabilities with the state-of-the-art tooling on such systems are also described.