Distribution Optimization Using Ant Colony Optimization (ACO) Method Case Research: PT. Coca Cola Official Distributor of Surabaya Area

PT.Coca-Cola Amatil Indonesia is a manufacturing company that produces beverages such as soft drinks, tea, milk, juice, isotonic and mineral water, located in Pandaan. This company has a distribution area that is spread almost all over Indonesia. In the Surabaya area, the company has 30 kiosks/salesmen as partners. The distribution of products in the Surabaya area is done through Take Order (TO) sales. The final product is distributed to consumers through third parties. The purpose of this research is to determine the distribution route with the shortest distance. The distribution problem experienced by this company is better known as the Traveling salesman problem (TSP). TSP is a combinatorial problem where when the problems faced are increasingly complex, the time needed is also getting longer. Several methods for solving TSP have been proposed. One of the best is the metaheuristic method, one of which is Ant Colony Optimization (ACO). In this research, the ACO method is used to solve the TSP problems encountered. The routes generated from the ACO method are 41.3 km which is 9.03% shorter than the actual route.


Introduction
The traveling salesman problem (TSP) aims to find the shortest travel route with several destination cities with certain lines where each city is only allowed visited once and the trip ends with returning to the original city (Lukman et al., 2011). TSP is categorized as an NP-hard problem with a high number of possible solutions (Halim & Ismail, 2019) which means that the minimum expected time to obtain optimal solution is exponential (Brezina & Čičková, 2011). Therefore, a metaheuristic approach is proposed to produce the best solution.
Metaheuristic methods have been widely used in previous researches. For some problems, this method can even provide optimal results with a comparatively shorter calculation time (Rahmawati & Santosa, 2017). Some of the problems that can be solved using this method include the problem of machine scheduling (Kundakcı & Kulak (2016) using the hybrid genetic algorithm method in the job shop scheduling problem, Utama et al. (2019) which uses cross entropy-genetic algorithm in flow shop sceduling problem to get minimal total tardiness, and also Ying and Lin (2020) who use simulated annealing for job shop scheduling without waiting time), project scheduling (Rahmawati & santosa (2016) that uses cross entropygenetic Algorithm for solving Resource-Constrained Project Scheduling Problem, Kadri and Boctor (2018) using genetic algorithm to solve the resource-constrained project scheduling problem with transfer times and also Lin et al. (2020) which used genetic programming hyper-heuristic approach to the multi-skill resource constrained project scheduling problem), as well as distribution problems (Erdianto et al. (2019) who used genetic algorithms and nearest neighbors to find distribution routes, Santosa et al. (2016)  249 used a hybrid cross entropy-genetic algorithm to solve multi-product inventory ship routing with a heterogeneous fleet model, Zang and Xiong (2018) that used ant colony optimization to find the best routes, and also Khadijah and Hasanah (2019) that used tabu search and differential evolution to find distribution routes) and many more.
In this research, the ant colony optimization (ACO) method which is part of Metaheuristic Methods is used to solve the distribution problems that occur. ACO was chosen because ACO was proven could produce the best route for distribution problems (Fahmi et al., 2020).

Research Method
The method used to solve the TSP problem in this research is the Ant Colony Optimization (ACO) method. Table 1 below is the destination location data. There are 34 demand nodes in the Surabaya area in November 2019. The first node in the table, CCOD Surabaya, is the origin node which is the location of the distribution center in Surabaya.  Table 2 below is a distribution route generated by the ACO algorithm. From this table, it is known that the first destination of the route is the Kopi Satu shop. While the last stall on the route is the Sams and Bakery shop following the sequence generated by the ACO algorithm. The total distribution distance generated by following the sequence obtained from the ACO algorithm is 41.3 km. The more iterations are generated, the more stable the results are obtained. From the results of running with the ACO algorithm (Figure 1), the results began to stabilize with a distance of 41.3 km on the 85th iteration.

Conclusion
The results obtained from this research are the Ant Colony Optimization algorithm produced a total delivery distance of 41.3 km whereas if using the actual route, the total delivery distance is 45.4 km. The Ant Colony Optimization algorithm method is proven to be able to save a distance of 4.1 km with a saving percentage of 9.03%.