The invention relates to improvement of a robot technology, and particularly relates to a method for stably controlling the gait of a biped robot under a slope condition. An improved DQN algorithm isobtained by replacing a comment function with a double-Q learning correction function, and the problem of walking control of the biped robot under the large slope condition is solved. The method comprises the steps of building a humanoid biped robot model, pre-training control parameters and biped robot walking motion stability. A simplified six-degree-of-freedom connecting rod model is adopted inthe humanoid biped robot model, and a planar foot bottom structure form is adopted in the feet. Controller training is carried out on the pre-training control parameters through deep reinforcement learning of an improved DQN network structure. The biped robot walking motion stability is achieved through a trained controller.
本发明涉及一种机器人技术的改进,具体为一种斜坡条件下双足机器人步态稳定控制方法,通过双Q学习修正函数取代评论函数得到一种改进的DQN算法,解决较大斜坡条件下双足机器人的步行控制问题包括建立仿人双足机器人模型、预训练控制参数和双足机器人步行运动稳定性;所述仿人双足机器人模型采用一种简化的6自由度连杆模型,足部采用平面足底结构形式;所述预训练控制参数利用改进的DQN网络结构的深度强化学习进行控制器训练;所述双足机器人步行运动的稳定性是通过训练的控制器实现。
Method for stably controlling gait of biped robot under slope condition
一种斜坡条件下双足机器人步态稳定控制方法
2020-10-02
Patent
Electronic Resource
Chinese
Gait control method of under-actuated biped walking robot
European Patent Office | 2021
|