Abstract: To overcome the drawback that the premature convergence is liable to take place in the process of optimal multicast tree searching by genetic algorithm, a chaotic genetic algorithm is proposed to inhibit the premature. Introducing the intrinsic stochastic property and ergodicity of chaos, the algorithm proposed employs the chaotic perturbation in GA to operate the population with a mutation-like chaotic perturbation when judging that the premature happens. Thus, the population becomes more diverse and the premature convergence can be overcome effectively with GA's ability in global search kept and algorithm performance greatly improved. Simulation results showed that the proposed algorithm can also build an optimal multicast tree efficiently and quickly to meet the requirement of QoS restriction.