This solver is very similar in principle (and code) with the existing K-means. It is a dynamic solver based on the principles outlined in [1], and adopting the Sweep method in [2].
Despite being old references, they are the basis of many recent approaches, and are also more practical for not requiring the knowledge of demand coordinates and accepting a continuous distribution of packages.
References
[1] Bertsimas, Dimitris J., and Garrett Van Ryzin. "Stochastic and dynamic
vehicle routing with general demand and interarrival time distributions."
Advances in Applied Probability (1993): 947-978.
[2] Gillett, Billy E., and Leland R. Miller. "A heuristic algorithm for the
vehicle-dispatch problem." Operations research 22.2 (1974): 340-349. NBR 6023
This solver is very similar in principle (and code) with the existing K-means. It is a dynamic solver based on the principles outlined in [1], and adopting the Sweep method in [2].
Despite being old references, they are the basis of many recent approaches, and are also more practical for not requiring the knowledge of demand coordinates and accepting a continuous distribution of packages.
References
[1] Bertsimas, Dimitris J., and Garrett Van Ryzin. "Stochastic and dynamic vehicle routing with general demand and interarrival time distributions." Advances in Applied Probability (1993): 947-978.
[2] Gillett, Billy E., and Leland R. Miller. "A heuristic algorithm for the vehicle-dispatch problem." Operations research 22.2 (1974): 340-349. NBR 6023