摘要
针对纯被动机器人对环境变化敏感,抗干扰能力差等问题,提出了一种基于Sarsa(λ)强化学习的底层PD控制器参数优化算法。在MatODE环境下建立双足有膝关节机器人模型并进行控制器设计。通过与传统控制器仿真结果的对比分析,得出该算法可使模型获得更加稳定的行走步态,同时提高了系统抵抗斜坡扰动的能力,增强机器人的行走鲁棒性。
For fully passive dynamic walking robot sensitive to the change of environment and poor in anti-interference,a parameters optimized algorithm for underlying PD controller based on the Sarsa(λ) reinforcement learning was proposed here.The robot model with knees and controller were built in the environment of MatODE.Compared with traditional controller we draw the conclusion that this algorithm can make robot get more stable gait,at the same time,improve the ability to overcome the slope disturbance and strengthen the walking robustness.
出处
《江南大学学报(自然科学版)》
CAS
2012年第2期132-136,共5页
Joural of Jiangnan University (Natural Science Edition)
基金
国家自然科学基金项目(60905049)
机器人技术与系统国家重点实验室(哈尔滨工业大学)自主课题项目(SKLRS200804C)
作者简介
臧希喆(1975-),男,黑龙江哈尔滨人,教授,硕士生导师。主要从事遥操作,被动机器人等研究。Email:zangxizhe@hit.edu.cn