SUNConferences, Computers and Industrial Engineering 42

Font Size: 
A Rich Multi-Period Delivery Vehicle Master Routing Problem Case Study
Angela Liza Rademeyer, Robert Benetto

Last modified: 2012-06-25


One of South Africa’s largest retail chains services 275 stores, each with a different average offload time, from their Gauteng distribution centre. Each store is visited on between 1 and 6 days of the 6-day week with some stores receiving more than one load per day. The objective of a study completed for Imperial Distribution was to assign stores to delivery profiles (permissible delivery day combinations) to meet pre-set daily workload targets. Such targets specified that the workload would be split according to specified ratios for each day of the week. Workload can be interpreted as; number of stores, total quantity delivered or total time worked. A rich vehicle routing problem (VRP) such as this one is usually characterised by the addition of real-world constraints to the standard academic problem, in this case the periodic VRP (PVRP), from which it is derived. This problem instance was solved using a memetic algorithm. While all stores were permitted to change nominated delivery days (NDDs) from their current assignments, some of the stores preferred not to change if it was avoidable. For these particular stores, their current NDDs were marked ‘preferred’ and the algorithm had an additional objective to maximise the total number of preferred day assignments. Once the NDDs had been found for an average/mid-week, low (reduced frequency) and peak (increased frequency) week scenarios were run. Days from existing profiles had to be removed/added respectively with minimal disturbance to the average schedule. The client was pleased with the quality of the results, which they had not been able to achieve with other methods, as well as the time taken to produce them.

Full Text: PDF

Comments On Paper

View all comments  |  Add comment