Physics-based motion synthesis is usually difficult and takes long computational time. However biological research works show that human beings an d animals take very little effort to control their motion. The idea is that instead of controlling mo tion in every detail, natural animals may only main tain or tweaks some qualitative properties of its motion and utilizes the complicate interaction between bo dy and environment. Inspired by this theory, in this p aper, a novel method is proposed for motion synthes is. Based on the theory of qualitative dynamics, adaptiv e motion control is achieved through manipulating the topological structure of the dynamic system, wh ich enhance the structural stability, rather than counteracting the perturbation effects. Compared wit h current methods, the new method is extremely efficient for it requires little computation and could be accelerated by GPU.