伸缩腿双足机器人强化学习步态生成方法  

Gait Generation Method for Biped Robot with Telescopic Legs Based on Reinforcement Learning

作　　者：崔文杰 唐俊 侯运锋 李清都 CUI Wenjie;TANG Jun;HOU Yunfeng;LI Qingdu(Institute of Machine Intelligence,University of Shanghai for Science and Technology,Shanghai 200093,China;School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学机器智能研究院,上海200093 [2]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《控制工程》2025年第2期280-289,共10页Control Engineering of China

基　　金：上海市浦江人才计划项目(2019PJD035);上海市人工智能创新发展专项项目(2019RGZN01041)。

摘　　要：近年来,相关研究已经证明能够通过强化学习生成稳健的双足机器人行走步态。然而,传统机器人的结构能效较低。针对这一问题,基于一种能效更高的伸缩腿结构的机器人,提出一种运用周期函数来交替奖励与惩罚双足触地的方法,来改进传统训练学习方法生成步态不规律的现象,鼓励智能体生成对称步态。实验证明,通过简单运用周期函数设计稠密奖励函数,可以有效指导伸缩腿机器人生成高效的运动模式,能够有效提高机器人的能效,且具有较高的鲁棒性。In recent years,related research has been able to generate robust bipedal robot walking gaits through reinforcement learning.To solve this problem,for a robot with a more energy-efficient telescopic leg structure,a method of alternately rewarding and punishing the double-foot touchdown using periodic functions is proposed to improve the phenomenon of irregular gait generation in traditional training and learning methods.Furthermore,it encourages the intelligent agent to generate a symmetrical gait.Experiments show that simply designing a dense reward function with a periodic function,can effectively guide the telescopic-legged robot to generate an efficient motion pattern,which can effectively improve the robot’s energy efficiency,and has high robustness.

关 键 词：强化学习 伸缩腿 双足机器人 步态生成 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]