摘要
在二级倒立摆控制器优化设计的研究中,由于二级倒立摆系统是一种复杂多变量、强耦合的不稳定系统,较难达到稳定平衡状态,首先要建立模型,之后才能够进行系统仿真与实际控制实验。由于系统对控制器性能要求较高,因此选用遗传算法训练的小波神经网络控制器,并针对遗传算法仍然存在的收敛速度慢,泛化性能差,可能陷入"早熟"等许多问题,对算法加以改进。将采用罚函数为基础的小生境技术引进到遗传算法中;并根据个体适应度来改进交叉概率。在仿真与实物控制实验中,控制器能够实现二级倒立摆系统的稳定控制,且抗干扰能力、系统平衡恢复速度优良,验证了设计的二级倒立摆控制器的有效性。
The double inverted pendulum system is a complex, multivariable, strong coupling system, which is hard to reach steady state. And the system requires higher performance to the controller. So genetic algorithm is proposed for constructing and training the wavelet network. And to solve a series of problems of genetic algorithm such as, slow convergence, bad generalization, poor calculation stability, and may gets into premature convergence and so on, we put forward an improved genetic algorithm. First, we join the niche technique into the genetic algorithm. Then, we use the improved adaptive crossover probability based on individual fitness value. In the system simulation and actual control experimental process, the controller can realize the stable control of the double inverted pendulum system, and the system has excellent capacity of resisting disturbance.
出处
《计算机仿真》
CSCD
北大核心
2015年第4期305-309,共5页
Computer Simulation
关键词
二级倒立摆
小波神经网络
改进遗传算法
Double inverted pendulum
Wavelet neural network
Improved genetic algorithm