基于Bloch量子遗传算法的倒立摆模糊控制器优化设计

摘要

To solve the optimization design of fuzzy controllers for inverted pendulums, an novel design optimization approach based on the Bloch quantum genetic algorithm ( BQGA) is proposed. This approach regards the three Bloch coordinates of each qubit as paratactic genes. Each chromosome contains three gene chains, and each of the gene chains represents an optimization solution that is a group of controller parameters, which can accelerate the convergence process for the same number of chromosomes as the common quantum genetic algorithm (CQGA). With the fuzzy neural network controller (FNNC) for a single inverted pendulum being an example, the experiment was perfomed and the control effect was discussed in detail based on two initial states. The experimental results demonstrate that the BQGA-based design is obviously superior to the CQGA-based design, and the FNNC designed using the BQGA-based approach is obviously superior to the LQR controller when the systematical parameter changes.%为解决倒立摆模糊控制器的优化设计问题,提出一种基于Bloch量子遗传算法(BQGA)的优化设计方案.该方案将量子位的3个Bloch坐标都看作基因位,每条染色体包含3条并列的基因链,每条基因链代表一个优化解,即一组控制器参数,在与普通量子遗传算法(CQGA)染色体数目相同时可加速优化进程.以模糊神经网络控制器(FNNC)的优化设计为例,以单级倒立摆为被控对象,针对两种初始状态,对控制效果进行了分析对比.实验结果表明,该方案优化的控制器明显优于基于普通量子遗传算法优化的同类控制器;当倒立摆系统参数改变时,该控制器也明显优于LQR控制器.