基于二级倒立摆的自适应神经网络模糊控制

Adaptive neural network fuzzy control for double inverted pendulum

针对二级倒立摆系统,提出了一种先进的智能控制策略.该种方法采用BP算法与梯度下降法结合的混合算法对Takagi-Sugeno模糊模型中的前项及后件参数进行优化修正,在已获得的客观输入输出数据对的基础上,提出一种基于自适应神经网络的模糊推理系统ANFIS来对倒立摆系统进行建模和控制.仿真结果表明,所提出的设计方法是正确的和可行的.

A novel intelligent control strategy based on double inverted pendulum is proposed. It is simple to be redized only by optimizing control parameters meanwhile using BP-GDA （backward propagation-gradient descent approach） hybrid algorithm. Based on captured objective input-output data couple, a new ANFIS （adaptive neural network-based fuzzy inference system） framework, which can construct fuzzy membership function and fuzzy rules in Takagi-Sugeno fuzzy model, is used to model and control the complex inverted pendulum. Simulation result shows that the method proposed is valid and applicable.