基于动态结构模型的两轮机器人平衡控制研究

Two Wheel Robot Balance Control Research Based on Dynamic Structure Model

摘要:

针对无监督学习网络模型对于特定的映射空间,期望产生特定的影响具有一定的局限性,提出了有监督的可生长结构的动态结构模型；该模型将动态结构模型与径向基函数网络结构有机结合,采用的是需要时生长的自组织生长算法,通过插入新的神经元、改变竞争层神经元间的连接以及调整层与层的连接权值,达到模型输出的精度要求,有效地解决了在输入特定的网络数据时,能产生期望的输出；该模型应用于两轮平衡机器人的平衡控制,经过仿真实验表明,该动态结构模型能够实现对两轮平衡机器人的平衡控制,且有一定的抗干扰性和实用价值。

Abstract:

For unsupervised learning network unsupervised learning network model for a specific mapping space, expect a specific impact has some limitations, supervised can grow structure network dynamic structure model. The model will combine unsupervised learning network model and radial basis function network architecture, uses a growth of self-organizing algorithm when necessary, by inserting a new connections between neurons, change the competitive layer neurons and adjust the weights between layer and layer, meet the requirements of the precision of the model output, effectually solved that when the input is specific network data, can produce the desired output. The model is applied to two rounds of the balancing of the robot control, through the simulation experiments show that, dynamic structure model model realized with two rounds of balance control of the robot, and has certain anti-interference and practical value.