Heuristics from nature for hard combinatorial optimization problems

Abstract

In this paper we try to describe the main characters of Heuristics ‘derived’ from Nature, a border area between Operations Research and Artificial Intelligence, with applications to graph optimization problems. These algorithms take inspiration from physics, biology, social sciences, and use a certain amount of repeated trials, given by one or more ‘agents’ operating with a mechanism of competition‐cooperation. Two introductory sections, devoted respectively to a presentation of some general concepts and to a tentative classification of Heuristics from Nature open the work. The paper is then composed of six review sections: each of them concerns a heuristic and its application to an NP‐hard combinatorial optimization problem. We consider the following topics: genetic algorithms with timetable problems, simulated annealing with dial‐a‐ride problems, sampling and clustering with communication spanning tree problems, tabu search with job‐shop‐scheduling problems, neural nets with location problems, ant system with travelling salesman and quadratic assignment problems.