Abstract-Variable-stiffness actuation exhibits promising features for obtaining human-like behavior and safer human robot physical interaction. Task planning and closed-loop control of these systems pose many challenges due to their complicated structure, and the need of satisfying many constraints during task execution. This paper introduces a framework for the design and numerical solution of time-optimal control problems for VSA systems. Two different time-optimal control problems, namely "minimum time for target performance" and "minimum time for maximum performance", are formally defined, and methods for solving them are presented based on existing numerical software tools for nonlinear optimization. Two experimental case studies, focusing on ball throwing tasks with antagonistically-actuated VSA systems, are used to test the presented methods and show their validity.