二级倒立摆的自适应神经网络控制

Adaptive neural network control for double inverted pendulum

倒立摆系统是一种典型的非线性、多变量、不稳定系统,目前,对于这种复杂对象的控制问题在控制领域具有十分重要的研究价值。针对此种非线性系统的控制问题,提出一种智能控制方法来解决这个问题。通过应用神经网络控制和模糊控制相结合的方式,集合二者的优点,提出一种将BP算法与最小二乘算法相结合的算法,对Takagi-Sugeno模糊推理系统中的参数进行优化修正,设计一种自适应神经网络的模糊推理系统来控制倒立摆,实验结果证明该理论是准确可行的,与LQR实时控制相比响应速度快、精度高。

The inverted pendulum system is characterized as a typical nonlinear, multi-variable, unstable system. At present, the research on the control of complex objects has great value in control area. This article is based on the control of the nonlinear system and put forward one intelligent method to solve this problem. This method combined neural network control and fuzzy control reflects the advantages of them. A mixed arithmetic of BP and LSE arithmetic are used, in order to optimize and amend the parameters of Takagi-Sugeno fuzzy model. A sort of self-adaptive neural-fuzzy inference system is put forward for control of inverted pendulum. The result of experiment shows this theory is feasible and effective. The response speed is faster than LQR control. And the precision is higher than it, too.