In this paper, we introduce the concept of non-dominance (commonly used in multiobjective optimization) as a way to incorporate constraints into the fitness function of a genetic algorithm. Each individual is assigned a rank based on its degree of dominance over the rest of the population. Feasible individuals are always ranked higher than infeasible ones, and the degree of constraint violation determines the rank among infeasible individuals. The proposed technique does not require fine tuning of factors like the traditional penalty function and uses a self-adaptation mechanism that avoids the traditional empirical adjustment of the main genetic operators (i.e., crossover and mutation).