摘要
构建了用于倒立摆平衡控制的神经网络学习模型。该模型利用可生长结构神经网络的优势,不需要预先规定网络的结构和规模,便可以在学习过程中根据需要生长。基于可生长结构的神经网络将监督与无监督学习结合,能够快速学习刺激与响应之间的潜在关系。该神经网络离线进行监督学习,训练后作为控制器作用于倒立摆系统,构成基于可生长结构的倒立摆控制模型。以Matlab为开发工具进行了仿真实验。仿真结果表明,该模型能够完成一级倒立摆平衡控制任务,并验证了其有效性和抗干扰能力。
A control model for balance of inverted pendulum is designed based on neural network.Takiug advantage of growing network, the model can grow rapidly during learning process, without predefining structure and size of neural network. Growing network combines supervised learning with unsupervised learning and gains underlying relation between stimulus and response. After supervised learning off line, trained neural network as a controller is applied to inverted pendulum system to form control model based on growing network. Simulation results in Mat- lab show that the model can accomplish single inverted pendulum balancing task, and has the anti-disturbance ability.
出处
《控制工程》
CSCD
2007年第5期485-487,共3页
Control Engineering of China
基金
国家自然科学基金资助项目(60375017)
教育部博士学科点专项科研基金资助项目(20050005002)
关键词
生长网络
倒立摆
自组织
监督学习
growing network
inverted pendulum
self-organiziug
supervised learning
