In this paper, three fundamental properties of a potential field-based flocking algorithm, i.e. merging of neighbouring graphs during the system evolution, collision avoidance and convergence of position of the centre of mass of informed agents to that of virtual leader are discussed. Next, these properties are utilized to determine required number of informed agents based on initial position of uninformed ones and consequently reduce the domain of search in optimization problems defined for finding the optimal number of required informed agents. Finally, a new optimization framework is proposed, which benefits Voronoi diagrams in order to reduce the number of informed agents required for velocity convergence of the whole group. This optimization framework reduces computational complexity in the cost of lower optimality.