Schedule optimization of freight vehicle fleet using data analytics

Abstract

Schedule optimization is a key decision process of fleet management. However, truck and driver scheduling in multi-plant goods distribution is a complex problem due to geographically distributed customer sites and plants, heterogeneity in trucks, driver behavior, varying traffic conditions, and constraints such as working and resting hours for drivers. Moreover, we need to satisfy conflicting objectives such as maximizing order coverage and minimizing of the overall costs. At present context, the scheduling process is typically handled by a fleet manager who is responsible for assigning both the trucks and drivers to meet the confirmed jobs/orders of a given day. Such scheduling usually happens on the evening of the day prior to the order delivery date. As an NP-complete problem, assigning most suitable pair of vehicle and driver while satisfying both company and customer becomes difficult in a situation where there is an increment of total number of orders. We propose an automated, heuristic-based truck and driver scheduling solution which comprises of a rule checker and a scheduler. Rule checker imposes constraints and conditions such as driver and truck availability, delivery time constraints, and operating and resting hours. A scheduler that applies simulated annealing is proposed to cover as many orders as possible while minimizing the overall cost. The utility of the proposed solution is tested using a workload derived from a real-world bulk-cement distribution company. The results show good coverage of orders where the coverage increased by more than 10% compared to manual scheduling while minimizing the total cost by 35%. Furthermore, the solution has flexibility to tolerate exceptions due to breakdowns, traffic congestion, and extreme weather conditions without a considerable impact on most of the already assigned pairs of vehicle and driver to orders.