Correction: Advantages of Task-Specific Multi-Objective Optimisation in Evolutionary Robotics

Many real-world optimization problems are evaluated in terms of multiple, often conflicting criteria or objective functions. When there is no a priori information about the importance of each objective, the solutions to such a multi-objective optimization (MOO) problem are usually compared in terms of Pareto dominance [1, 2]: A solution dominates another one if the former is not worse than the latter in all objectives and strictly better in at least one. The goal when tackling such a MOO problem is to find, or approximate as well as possible, the set of all solutions whose image in the objective space is not dominated by any other feasible solution. This set is called the Pareto set and its image is called the Pareto front. Computing the Pareto front is often intractable in practice and heuristic methods are necessary to generate a high-quality approximation [3]. Among the heuristic methods, multi-objective evolutionary algorithms (MOEAs) have achieved a considerable success and we refer the reader to the many textbooks available on the subject for a detailed introduction [1, 2].