基于遗传算法的神经网络预测控制及应用

【作者】 梁建辉；

【导师】 安锦文；

【作者基本信息】 西北工业大学 ， 导航、制导与控制， 2003， 硕士

【摘要】 本文把预测控制与神经网络、遗传算法等智能技术相结合，对其在无人机中的应用进行了设计和研究。 预测控制是上世纪七十年代发展起来的一种工业高等过程优化控制方法。它具有建模简单方便、鲁棒性好等特点。它对模型要求不高，却具有较高的控制性能。工业中应用最为普遍的预测控制大都是基于线性渐进稳定的系统设计出来的，算法虽然简单，但应用受到一定的限制。对于无人机这样的典型非线性系统，预测模型必须采用非线性模型。神经网络在非线性建模方面具有独特优势。本文利用神经网络作为预测模型，以遗传算法作为在线优化算法，对无人机的高度爬升进行了仿真研究。同时，根据遗传算法强大的随机全局搜索机制，对其在神经网络的训练方面进行了研究。更多还原

【Abstract】 This paper studied on the combination of predictive control, artificial neural network and genetic algorithm, and applied it to unmanned vehicles.Predictive control is an optimizing control algorithm which developed in 1970s. It has the advantage of convenience of modeling and good robust characteristic. With simple or low-precise predictive model, this control strategy can let control system gain high control quality. However, applied predictive control algorithm mostly is based on linear stable system. The nonlinearity of unmanned vehicles requires a nonlinear model in predictive control. It is convenient to use artificial neural network to model such a nonlinear system. This paper used neural network as the predictive model, and genetic algorithm as optimizing algorithm, to simulate the unmanned vehicle’s ascending process. In addition, this paper also studied on optimizing neural networks utilizing the global searching mechanism of genetic algorithm.更多还原