Abstract: The objective of this study is to obtain dynamic stable cycle gait for underactuated biped robot.The miscellaneous dynamic model of the robot is built first and then gait is planned by time-invariant strategy.Since the stability of the cycle gait is determined by its parameters，genetic algorithm is introduced to optimize them and using optimal moment as the goal and restrictive conditions as the punish functions.Virtual prototype is used to simulate the walking process and verifies the gait.Simulation shows that joints movement convergence to stable limit cycles，planned walking pattern is realized，and feasibility of the proposed method is verified.

Key words: underactuated biped robot, gait planning, genetic algorithm, time-invariant gait

摘要： 以欠驱动双足机器人为对象研究其周期稳定的动态步态规划方法。首先建立欠驱动双足机器人的混杂动力学模型，然后采用时不变步态规划策略对机器人步态进行规划，并研究周期步态的收敛条件。步态参数直接决定周期步态的稳定性，采用遗传算法，以能耗最优为目标，以限制条件为约束对步态参数进行选择和优化。最后通过虚拟样机对机器人的行走过程进行动力学仿真。实验表明规划步态收敛于稳定的极限环，实现了高速动态步行，该规划方法是可行的。

关键词: 欠驱动双足机器人, 步态规划, 遗传算法, 时不变步态