Data learning based hypersonic flight control using ELM

This paper is towards the controller design using extreme learning machine (ELM) for the longitudinal dynamics of a generic hypersonic flight vehicle (HFV). The basic idea is to train the data learning from previous controller and then obtain the optimal weight. In the first step, the existed the back-stepping controller with high order neural networks (HONNs) is borrowed to collect the required data. The “adaptive behavior” of existed the back-stepping controller is trained and tested by batch learning of ELM. Then the optimal parameters obtained from ELM are used as initialization to construct the feedback design for controller. In this way, the prior information of nominal design is not needed and there is no need of online learning for the neural networks (NNs). The simulation study is presented to show the effectiveness of the proposed control approach.