自适应混沌蚁群算法的神经滑模控制器仿真

RBF Neural Sliding Mode Control Simulation Based on Adaptive Chaotic Ant Colony Method

针对RBF神经网络权值在线优化较慢的问题,通过自适应混沌蚁群算法对神经网络的权值进行离线优化,利用在线自适应算法对神经网络的权值进行局部调整,提出了一种自适应混沌蚁群神经网络学习方法,该方法缩短了神经网络的在线学习时间,同时也抑制了滑模的抖振。仿真结果表明,基于自适应混沌蚁群算法的神经滑模控制具有强的鲁棒性,完全适用于不确定干扰上界未知的复合试验系统的振动控制,并且降低了系统的保守性。

Due to the problem of slow online optimization for weights of RBF neural network, an adaptive chaotic ant colony RBF neural learning method is proposed, which optimizes neural network weights offline by adaptive chaos ant colony firstly and then adjust local weights online using adapting method. The method shortens the online neural network learning time and effectively restrains the chattering of sliding mode. The algorithm has strong robustness and is suitable for the vibration control of the composite test system with unknown disturbance, and reduces the conservatism of the system.