基于补偿滑模神经网络的某炮控系统位置控制

Gun control system based on self-organizing neural network with complementary sliding modes

针对某炮控系统存在较强的非线性和不确定性特征,提出了基于补偿滑模的自组织神经网络控制策略。引入了补偿滑模面设计方法,构成了自组织神经网络控制器和辅助补偿器。自组织神经网络控制器由Hermite多项式、变结构神经网络和神经元参数自学习算法构成,其减小了计算复杂度,提高了自适应能力;梯度下降法对神经网络的参数进行自学习,提高了系统的收敛速度;辅助补偿器的引入进一步减小了系统稳态误差,满足了该炮控系统的基本指标要求,保证了系统在Lyapunov意义下的稳定性和鲁棒性。半实物仿真试验表明:该控制策略有效地提高了系统的控制精确度和鲁棒性,减小了外界干扰对系统性能的影响。

A self-organizing neural network with complementary sliding modes control strategy is proposed for the strong nonlinearities and uncertainties of a gun control system(GCS),which consists of the self-organizing neural network controller(SNNC)and the auxiliary compensation controller(ACC)with the complementary sliding mode surface.The self-organizing neural network controller included a Hermite polynomial,a variable structure self-organizing neural network(VSSNN)and self-learning parameters with the gradient descent method,which reduced the computational complexity and accelerated the ability of adaptation.The gradient descent method adjusted parameters of the neural network and promoted the convergence rapidity.The auxiliary compensator was introduced to further reduce steady-state error of the system,which satisfied the basic indicators of requirements and guaranteed the stability and robustness of the system in the sense of Lyapunov.The semi-physical test simulation shows that the control strategy greatly improves the control accuracy and robustness of the system,and effectively eliminates the influence of disturbance in the system.