Résumé : Combinatorial optimization problems can be found in many aspects ofmanufacturing, computer science, logistics and many more. These problemsconsist in combining a finite set of elements so that a cost measure isminimized or a quality measure is maximized. Despite the great interestgenerated by the many practical applications, combinatorial optimizationproblems can be quite hard to solve. In fact, many combinatorial optimizationproblems, like the traveling salesman problem and the permutation flowshopproblem, belong to a class of problems called NP-hard. The techniques used tosolve these problems can be grouped in two classes, exact methods andapproximate methods. Exact methods are guaranteed to eventually find theoptimal solution. Yet, the time needed to find the optimal solution may beimpractical. On the contrary, approximate methods are not guaranteed to findthe optimal solution but, in most cases, can find solutions with a qualityclose to the optimal in little time. Among these methods, stichastic localsearch (SLS) algorithms have been proved to be very successful. In fact, SLSalgorithms comprehend many of the most widely known high performance algorithmsto solve hard combinatorial optimization problems. These algorithms are oftenobtained after a significant, manual algorithm engineering effort. It ispossible to automatize this process by using automatic configuration tools witha configurable algorithmic framework. Such frameworks implement one or more SLSalgorithms in a modular way, where an algorithm is composed of differentalgorithmic components. This process is called automatic algorithm design(AAD). In this thesis, we expand the work done on grammar based automaticdesign of stochastic local search algorithms. In particular, we present a newalgorithmic framework, EMILI. This new framework improves over previousframeworks thanks to its modular design and its ability to instantiatealgorithms at run time. Using AAD, we present new state-of- the-art algorithmsfor the major objectives of the permutation flowshop problem (PFSP) and PFSPvariants with additional constraints. While working on PFSP we introduced a newspeed-up mechanism for the calculation of the objective function for PFSP withthe weighted tardiness objective as well as a new state-of-the-art algorithmfor PFSP with the makespan objective. Finally, we analyze how algorithmcomplexity affects the performances of automatically generated SLS algorithms.