基于神经网络旋转二级倒立摆系统的预测控制

Predictive Control of a Double Inverted Pendulum Based on Neural Network

针对倒立摆的研究,提出了一种应用神经网络和预测控制相结合的算法,用于控制旋转二级倒立摆系统。为了提高跟踪精度和快速性,可以把线性控制的条件作为非线性最优控制性能指标的约束条件,并将一种新型的非线性混沌映射引入到神经网络的参数学习算法中实现其权值调节,以逐步建立被控对象合理的多步预测模型。仿真结果表明,神经网络预测控制算法具有响应速度快、控制效果好和跟踪精度高等特点。可使非线性系统的预测问题得到较好的解决。

A control method combining neural network with predictive control is proposed,which controls a double inverted pendulum system.A new type of nonlinear chaotic map was introduced into the learning algorithm of neural network parameters in order to realizing the weight regulation.The weight of the network was modified to create a reasonable multi-step predictive model.It is shown by the simulation results that this algorithm has the characteristics of rapid response,good control quality and high tracking accuracy.