Solution for TSP/mTSP with an Improved Parallel Clustering and Elitist ACO


KARATAŞ BAYDOĞMUŞ G.

Computer Science and Information Systems, cilt.29, sa.1, ss.195-214, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 29 Sayı: 1
  • Basım Tarihi: 2023
  • Doi Numarası: 10.2298/csis220820053b
  • Dergi Adı: Computer Science and Information Systems
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Applied Science & Technology Source, Computer & Applied Sciences, INSPEC, Directory of Open Access Journals
  • Sayfa Sayıları: ss.195-214
  • Anahtar Kelimeler: ACO, Parallel ACO, Parallel Kmeans, TSP, ANT COLONY OPTIMIZATION, ALGORITHM, SYSTEM
  • Marmara Üniversitesi Adresli: Evet

Özet

Many problems that were considered complex and unsolvable have started to solve and new technologies have emerged through to the development of GPU technology. Solutions have established for NP-Complete and NP-Hard problems with the acceleration of studies in the field of artificial intelligence, which are very interesting for both mathematicians and computer scientists. The most striking one among such problems is the Traveling Salesman Problem in recent years. This problem has solved by artificial intelligence’s metaheuristic algorithms such as Genetic algorithm and Ant Colony optimization. However, researchers are always looking for a better solution. In this study, it is aimed to design a low-cost and optimized algorithm for Traveling Salesman Problem by using GPU parallelization, Machine Learning, and Artificial Intelligence approaches. In this manner, the proposed algorithm consists of three stages; Cluster the points in the given dataset with K-means clustering, find the shortest path with Ant Colony in each of the clusters, and con-nect each cluster at the closest point to the other. These three stages were carried out by parallel programming. The most obvious difference of the study from those found in the literature is that it performs all calculations on the GPU by using Elitist Ant Colony Optimization. For the experimental results, examinations were carried out on a wide variety of datasets in TSPLIB and it was seen that the proposed parallel KMeans-Elitist Ant Colony approach increased the performance by 30% compared to its counterparts.