Discrete online TSP

In this paper we introduce a discrete version of the online traveling salesman problem (DOLTSP). We represent the metric space using a weighted graph, where the server is allowed to modify its route only at the vertices. This limitation directly affects the capacity of the server to react and increases the risk related to each decision. We prove lower bounds on the performance of deterministic online algorithms in different scenarios of DOLTSP, and we present distinct algorithms for the problem, some of them achieving the best possible performance. We measure the performance of the algorithms using competitive analysis, the most widely accepted method for evaluating online algorithms. Besides, we perform an empirical simulation on paths, generating a significant set of instances and measuring the quality of the solutions given by each algorithm. Our experiments show that algorithms with the best competitive ratio do not have the best performance in practice. Key words: TSP, discrete me...

• Type: Journal
• Year: 2009
• Where: AAIM
• Authors: Mauro Aprea, Esteban Feuerstein, Gustavo Sadovoy, Alejandro Strejilevich de Loma
• Keywords: AAIM 2009, Algorithms, Metric Spaces, Online