This thesis investigates the role of duality and the use of approximation methods in cooperative optimization and control. Concerning cooperative optimization, a general algorithm for convex optimization in networks with asynchronous communication is presented. Based on the idea of polyhedral approximations, a family of distributed algorithms is developed to solve a variety of distributed decision problems, ranging from semi-definite and robust optimization problems up to distributed model predictive control. Optimization theory, and in particular duality theory, are shown to be central elements also in cooperative control. This thesis establishes an intimate relation between passivity-based cooperative control and network optimization theory. The presented results provide a complete duality theory for passivity-based cooperative control and lead the way to novel analysis tools for complex dynamic phenomena. In this way, this thesis presents theoretical insights and algorithmic approaches for cooperative optimization and control, and emphasizes the role of convexity and duality in this field.