提出了一种基于学习的动力学前馈方法 ,解决了欠驱动系统由于系统不完全可控造成的反馈控制失效、运动轨道难以实现的问题 .在此基础上提出了学习 -优化交替式最优运动轨道生成方法 ,该方法将轨道优化与学习控制相结合 ,所生成轨道可以在欠驱动机器人上实际操作 .对一台实际的欠驱动机器人实现了最优轨道生成 ,并进行了动作实验 ,取得了较好效果 A dynamic feedforward method based on learning is proposed to solve the problem that the under-driven system can not realize the feedback control due to incomplete control of the system and the orbit of motion is difficult to be realized. On the basis of this, a learning-optimizing alternating optimal motion Orbit generation method, the method combines orbit optimization with learning control, and the generated orbit can be operated on the under-driven robot.Optimum orbit generation is realized for an actual under-driven robot, Better effect