摘要
针对导弹拦截问题,提出一种自适应RBF神经网络滑模制导律.首先根据准平行接近原理和变结构控制理论设计滑模面,然后将滑模面作为RBF神经网络的输入变量,输出量即为导弹的加速度.为了使得导弹系统能够到达滑模面,采用自适应算法实时在线调整RBF神经网络的连接权值.该导引律将目标机动视为干扰量,在拦截过程中不需要测量目标加速度,因此该导引律对目标机动具有较强的鲁棒性.在执行上,只用到了视线角速率,因而实现简单.仿真结果表明,所提出的导引律和比例导引相比在脱靶量、拦截时间等方面有了很大的提高.
A new adaptive radial basis function neural network （RBFNN） sliding mode guidance law was proposed for intercepting maneuvering targets. First of all, we designed a sliding-surface using a quasi-parallel approach principle and variable structure control theory. We then used the sliding surface to input variables to the RBF neural network. In this case, the output was missile acceleration. In order to place the missile system on the sliding surface, we employed an adaptive algorithm that adjusts in real-time the connection weights of the RBF neural network. The missile acceleration in a given direction was determined by considering the target＇s acceleration as a disturbance, and thus it was not necessary to measure the target＇s acceleration directly. Therefore, this guidance law has strong robustness to target maneuvering. The new guidance law, which utilizes line-of-sight （LOS） measurement only, is simple to implement. Numerical simulations showed that the proposed guidance law yields better performance than proportional navigation.
出处
《智能系统学报》
2009年第4期339-344,共6页
CAAI Transactions on Intelligent Systems
基金
国家自然科学基金资助项目(60773065)
