Vehicle Routing Problem in Post Office Delivery for the posts
Transportation cost accounts for approximately half of the total operational expense for a logistics entity such that it should possess decent solutions to manage transportation to reduce the cost. A post office is the logistics establishment that has to manage a group of delivery vehicles where its current routing solution might not be effective, relying on skills of delivery drivers which could lead to unnecessary consumption of fuel; thus, determining the optimal vehicle routing to deliver parcels, known as Vehicle Routing Problem (VRP), could help save its cost greatly. This study used Evolutionary-based algorithms, Differential Evolution (DE) combined with Large Neighborhood Search (LNS) to find optimal routing for delivery vehicles belonging to the Chiang Rai post office, Thailand. The approach was applied to the routing of two vehicles responsible for one delivery zone in urban Chiang Rai area. The results showed that such optimization method had the average total distance per day per vehicle of 19.43 km, ~32.7% lower than that of current routing employed by the post office, given that 29 days of data for vehicle routing were tracked. The proposed combined method could be used as an effective decision-maker in the routing process for the post office. Vehicle Routing Problem in Post Office Delivery for the posts
Management of vehicle routing in an organization that is involved in transportation operation is a daunting task, especially if the survival of such entity largely depend on how it handles the routing complexity. For industries related to logistics, the large portion of operational cost is from the
transportation activity [1] such as fuel expense and driver wages. Therefore, a logistics establishment should determine the optimized ways to deal with the routing problem effectively i.e. having an exceptional solution to the transportation management such that it could drastically
minimize its overall operation cost, saving both time and expenses as well as satisfying its customers.
The post office in focus is the Chiang Rai post office branch, situated on Uttarakit Road, Tumbon
Viang, Chiang Rai Province, Thailand. This particular post office branch is responsible for both receiving and distributing parcels in the areas of Tumbon Viang, Rob- Viang, Mae Korn, Sansai, Mae Yao and Rimkok. It has the total of sixteen delivery vehicles for both normal and EMS
(Express Mail Service) operations; in those sixteen vehicles, there are four trucks and fourteen motorcycles. Each operational day, as parcels enter the post office, they are handled and distributed to customers at addresses in various areas in a timely manner; the parcels are assigned to drivers
of delivery vehicles according to their responsible zones, with some parcels requiring an expedited delivery process. Once the post office designates and loads packages to the delivery vehicles, the drivers would use their familiarity and instinct for the routing processes for the delivery; the
exercise of mastery for delivery in the areas is one of the major problems of the post office operation. The instinctive routing method might not be efficient enough, causing slow delivery and high fuel consumption; additionally, if new drivers were assigned to an area not familiar to them, the
delivery could face a major delay or could not match assigned addresse
The delivery routes of both vehicles were tracked by GPS-enabled smartphones; an application was installed to retrieve the routing information and total distance traveled as well as times and location coordinates; the application recorded a coordinate whenever the vehicle stops for a brief period (more than 3 seconds). The stored coordinates were then mapped onto the map application (Google Earth Pro®) to show the exact physical locations; the visual presentation also helped verify if the coordinates were correct. Next, the coordinate data were filtered and arranged such that some undesired coordinates, e.g. stops at traffic lights and other non-registered addresses, were removed. Then, a python
script determined the shortest distances between each pair of coordinates via the googlemaps module (Google Map API), and automatically constructed a distance matrix that has 𝑛 × 𝑛 dimension where 𝑛 equals to the number of travel coordinates including the main post office. Noting that this type of matrix is not a symmetric one because the shortest distance from point 𝐴 to point 𝐵 might not be the same as
that from point 𝐵 to point 𝐴 considering some routes might enforce a one-way traffic
There would be some issues if this method were going to be used for a real routing application; those issues refer to how the addresses on parcel can be converted into GPS coordinates and unexpectedly increased run-time of the optimization program when the number of delivered addressed is large. And number of delivery vehicles can affect program run-time. To make the entire process truly practical, first, decent data management of addresses must be arranged and image processing components combined with artificial intelligence (AI) software must be installed to automatically map customer addresses into GPS coordinates such that the optimization program could readily transform
the coordinates into distance matrices which are crucial elements to the entire process. In addition, certain parameters must be programmatically adjusted, depending on number of customer addresses to deliver and number of vehicles used for delivery, to make the program run-time within acceptable range of the post office operation.