3关节单杠体操机器人的动力学参数辨识

在采用拉格朗日方法确定了3关节单杠体操机器人动力学模型结构的情况下,动力学参数的精确辨识对机器人实时控制的实现显得十分重要.为实现对体操机器人多个动力学参数的精确辨识,在传统的遗传算法中,通过引入混合编码、海明距离、可变精度的交叉操作、正交试验设计、动态编码和反馈式突变等思想,再加上特殊设计的适应度函数,形成了一种改进的遗传算法.该算法在统计上更加合理,鲁棒性更强,更容易搜索到接近全局最优的可行解.通过体操机器人各个关节自由运动实验与模型数值仿真实验数据的实际比较,验证了所提出改进遗传算法的有效性,实现了3关节单杠体操机器人模型更为优化的动力学参数辨识.

Dynamic parameter identification of three-link acrobot on horizontal bar

In modeling the three-link acrobot on horizontal bar using Lagrange equations,the accuracy of dynamicparameters plays an important role in the realization of a real-time control.In this paper,an improved genetic algorithm(IGA) is proposed for parameter identification.The IGA introduces some new operators such as hybrid encoding,variable-precision crossover,integration ideas of orthogonal experiment design,Hamming distance,dynamic encoding and large-size mutation with feedback.In addition,a novel fitness function is used to evaluate the solutions.Therefore,the IGAcan be more robust,statistically rational,and easier in obtaining the solutions.Finally,by comparison of data betweenfree response and numerical simulation of each joint,the IGA is shown to be effective in the determination of dynamicparameters of three-link acrobot on horizontal bar.