This manuscript presents the research activities I conducted as an Associate Professor (« Maître de Conférences ») with the University of Lille since 2010. They deal with search heuristics for black-box multi-objective optimization, and they are articulated along three complementary research lines. Firstly, we consider landscape analysis as a central concept for understanding the foundations and behavior of multi-objective search heuristics. A number of general-purpose landscape features are proposed and analyzed for characterizing multi-objective landscapes. They allow us to better understand the difficulties that algorithms have to face depending on the problem being solved. They are subsequently used to predict algorithm performance and to automate the choice of which algorithm to select for solving a previously-unseen problem. Secondly, given that multi-objective optimization aims at identifying a set of solutions, it becomes relevant to consider the search space as the collection of all feasible sets of solutions. We start by clarifying the differences and similarities between sets according to different set preference relations from the literature. We further specify a set-based multi-objective local search, and we investigate the search difficulty as a function of the problem characteristics and of the considered set preference relation. At last, we contribute to the design and the improvement of efficient multi-objective search approaches. To this end, we rely on the concept of decomposition, that consists in decomposing the considered multi-objective optimization problem into a number of scalar sub-problems that are solved concurrently and cooperatively. This allows us to propose a number of distributed approaches that incorporate a high level of parallelism, and that can be deployed on modern computing environments. We also consider surrogate models to the evaluation function, and we investigate their integration into the multi-objective search process in order to address particularly expensive problems. We conclude the manuscript with some perspectives for massive optimization.