Most impedance control schemes in robotics implement a desired passive impedance, allowing for stable interaction between the controlled robot and the environment. However, there is little guidance on the selection of the desired impedance. In general, finding the best stiffness and damping parameters is a challenging task. This paper contributes to this problem by connecting impedance control to robust control, with the goal of shaping the robot performances via feedback. We provide a method based on linear matrix inequalities with sparsity constraints to derive impedance controllers that satisfy a H-infinity performance criterion. Our controller guarantees passivity of the controlled robot and local performances near key poses.