A fuzzy controller is usually designed by formulating the knowledge of a human expert into a set of linguistic variables and fuzzy rules. One of the most successful methods to automate the fuzzy controllers development process are evolutionary algorithms. In this work, we propose a so-called ``approximative'' representation for fuzzy systems, where the antecedent of the rules are determined by a multivariate membership function defined in terms of Voronoi regions. Such representation guarantees the $\epsilon$-completeness property and provides a synergistic relation between the rules. An evolutionary algorithm based on this representation can evolve all the components of the fuzzy system, and due to the properties of the representation, the algorithm (1) can benefit from the use of geometric genetic operators, (2) does not need genetic repair algorithms, (3) guarantees the completeness property and (4) can implement previous knowledge in a simple way by using adaptive a priori rules. The proposed representation is evaluated on an obstacle avoidance problem with a simulated mobile robot.