A method is presented to predict phase relationships between coupled phase oscillators. As an illustration of how the method can be applied, a distributed Central Pattern Generator (CPG) model based on amplitude controlled phase oscillators is presented. Representative results of numerical integration of the CPG model are presented to illustrate its excellent properties in terms of transition speeds, robustness and independence on initial conditions. A particularly interesting feature of the CPG is the possibility to switch between different stable gaits by varying a single parameter. These characteristics make the CPG model an interesting solution for the decentralized control of multi-legged robots. The approach is discussed in the more general framework of coupled nonlinear systems, and design tools for nonlinear distributed control schemes applicable to Information technology and robotics.