Vehicle Route Optimization for Ethiopian Pharmaceutical Delivery Using Reinforcement Learning Model

Abstract

This study considers the route optimization of EPSS pharmaceutical delivery in Ethiopia. The agency has 19 hubs across the country. Due to its centralized distribution of pharmaceutical products from a single hub in the cluster, it is experiencing issues as the number of facilities throughout the cluster rapidly increases. The goal is to lower transportation costs while maintaining a high degree of customer satisfaction. As a result, the focus is on the vehicle routing problem (VRP) within each facility of this enormous distribution chain. This study uses real data from the EPSS catchment area. One of the efforts to improve the quality of service is to provide optimization of the distribution process. Optimization can be done by adding a clustering approach to the existing reinforcement learning algorithm. The large coverage distribution by one vehicle takes more time and more cost. In the process of optimizing the route facility clustering method which has produced the clustered facilities. Then the clustered facilities are inserted into the RL in order to provide the best-optimized route. Furthermore, the delivery cost will be calculated based on the distance covered by the vehicle based on the optimal route. The results of this study are examined with the proposed 3 types of instances which are VRP20, VRP50, and VRP100. The researcher performed a statistical analysis of the results from various aspects and designed two experiments to evaluate. Finally, the researcher concludes the applicability approach for the selected type of problem and suggest directions in which researchers can improve the route optimization approach.