Evolutionary solutions to a highly constrained combinatorial problem

Abstract

Scheduling under constraints is a NP-problem which is found in many practical applications such as the job shop scheduling and the construction of the time table for a public transportation system or for the educational courses of a school. However, many sub-optimal algorithms have been developed for this problem, starting from different approaches going from the more classical ones proposed by operational research and graph theory to evolutive algorithms. Three evolutive algorithms: a simple genetic algorithm (D.E. Goldberg, 1989); a complex genetic algorithm (A. Colorni et al., 1990); and stochastic hill climbing (T. Back, 1991 and M. Herdy, 1990) are compared and evaluated on a particular instance of the time table problem. The selected test case consists of constructing the time table for a school where a set 6 constraints must be simultaneously satisfied.<>