基于惩罚函数的多模态平稳过渡策略研究

Research on multi-modal smooth transition strategy based on penalty function

为了使智能控制中多模态之间能实现平稳过渡,提出一种基于惩罚函数的多模态控制策略.根据系统特性及其不同的工况条件,在控制的母空间中划分多个子空间,每个子空间对应一个模态,每个模态采取不同的有针对性的控制策略,并把与每个模态相对应的控制策略推广到母空间,在每个子空间中对应控制策略的控制值所含权值较大,推广到别的子空间时控制值所含权值都受到惩罚.试验结果表明,用基于惩罚函数的多模态控制策略对纳米级驱动部件进行控制,当进给力为5 000 N时,控制精度达到10 nm.

Multi-modal control strategy based on penalty function was proposed to realize mode smooth transition in intelligent control. According to system characteristics and different operating terms, many sub spaces among mother spaces was divided, each sub space corresponding to a mode, and different pointed control strategy of each mode was adopted. Then the control strategy corresponding to each mode was popularized to the mother space, with the corresponding control strategy including relatively big correct value in each sub space, and the correct value was punished while popularizing to other sub space. The results of experiments showed that high control precision of about 10 nm can be obtained by using the multimodal control strategy based on penalty function when the feeding force was 5 000 N.