In this thesis we investigate the dynamic detection of symmetry relations in combinatorial prob-lems modeled as constraint satisfaction problems (CSPs). We examine how to exploit these sym-metries in order to generate a compact representation of the solution space and overcome the com-plexity barrier that undermines the efficient solving of these problems. We assess the combination of these techniques with the best known strategies for improving the performance of search such as dynamic ordering heuristics and full lookahead strategies. We demonstrate that the benefits drawn from our approach are orthogonal to, and benefit from, such combinations. We thoroughly validate these improvements through both theoretical and empirical means. Our experiments show the utility of dynamic symmetry detection on a full range of problems (i.e., from easy to difficult, for toy, real-world, and randomly generated problems). We demonstrate that these techniques are useful when finding one and all solutions, under static and dynamic ordering heuris-tics, and using partial and full lookahead strategies. In doing so, we dispel common notions that the dynamic computation of symmetry is too costly to be of practical utility. We also establish that the dynamic detection and exploitation of symmetries is a powerful, cost-effective tool for dramatically reducing the peak of the phase transition, possibly the most critical phenomenon challenging the efficient processing of combinatorial problems in practice. Although most of our work focuses on binary CSPs, we show how it can be extended to non-binary problems. We focus on the computational aspects of symmetry detection, and identify direc-tions for future research and their impact on other disciplines, such as AI Planning, visualization, and relational databases.