The generation of behavior for autonomous robots in complex environments is restricted by a potentially large number of constraints that must be observed. For many tasks, the relevant constraints can be expressed in a specific, low-dimensional reference frame. The dynamic systems approach to autonomous robotics represents constraints as attractors or repellors over such behaviorally relevant variables. This article presents a framework in which each constraint is treated independently as a dynamical system over a behavioral variable and provides a mechanism for the integration of these dynamical systems as contributions to a common vector field for movement generation. As a demonstration, we treat manipulation tasks for long cylindrical objects as a collection of separate constraints that are interpreted and solved with the proposed techniques. The resulting system is implemented on the small humanoid robot Nao and used to solve two exemplary movement tasks.