An improved version of the evolutionary Non-dominated Sorting Genetic Algorithm II (NSGA-Ⅱ) is used for this multi-objective optimization problem. The random number generators in the stochastic algorithm are replaced by a discrete time chaotic map for greater effectiveness. The Pareto optimal fronts, showing the trade-offs between different design objectives are obtained for the PI~λD~μ controller to illustrate the available choices to the designer.