一种基于优势更新的机器人平衡控制算法  

A Balance Control Algorithm of the Robot Based on Advanced Updating

作　　者：史涛[1] 任红格[1] 

机构地区：[1]河北联合大学电气工程学院,河北唐山063009

出　　处：《山东科技大学学报（自然科学版）》2013年第3期17-21,共5页Journal of Shandong University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(61203343)

摘　　要：针对自平衡机器人运动平衡控制问题,提出了一种基于优势更新的强化学习机制作为机器人的自平衡仿生学习算法。该算法利用优势更新中的基线,结合强化学习中的概率好奇心机制,以一定的概率选择优等行为,剔除劣等行为,从而使机器人在未知环境下可获得像生物一样的自主学习技能,实现机器人的仿生自平衡运动控制。最后,应用该算法对机器人进行自平衡的仿真实验。结果表明,这种基于优势更新的强化学习算法能使机器人获得较强的平衡控制技能,取得了较好的动态性能,体现了机器人的仿生特性。Aiming at the movement balance control problem of the self-balance robot, the reinforcement learning mechanism based on the advanced updating was proposed as a self-balance bionic learning algorithm of the robot. This algorithm can choose optimal behavior with the certain probability by using baseline in the advanced updating and combining the probability curiosity mechanism in the reinforcement learning,and get rid of the inferior behavior, so that the robot can obtain the bionic self-learning skills like creature under the unknown environment, and realize the bionic self-balance control of the robot. Finally,the simulation experiment on robot self-balance control was made by use of the bionic learning algorithm,and the result indicates that the robot can obtain the stronger self-learning control skills and gain tbe better dynamic performance by applying the reinforcement learning algorithm based on the advanced updating;in addition, the bionic characteristic of the robot is embodied.

关 键 词：优势更新 强化学习 好奇心 仿生 机器人 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]