A Benchmark Generator for Dynamic Optimization

Abstract:- Almost all real-world problems are dynamic and as such not all problem instances are known a priori. Many strategies to deal with dynamic problems exist in the literature, however they are not fully tested on combinatorial problems due to the deficiency of benchmarks. This paper presents a benchmark generator that uses Genetic Algorithms to produce benchmark problems for the dynamic traveling salesman problem. Various strategies of solving dynamic problems are compared on the new benchmarks. Generality of the Genetic Algorithm was retained so as it can be easily adapted for other kinds of combinatorial problems. Key-Words:- dynamic optimization, traveling salesman problem, genetic algorithms, combinatorial problems