四旋翼的改进PSO-RBF神经网络自适应滑模控制  

Improved PSO-RBF neural network adaptive sliding mode control for quadrotor systems

作　　者：唐志勇[1] 马福源 裴忠才[1] TANG Zhiyong;MA Fuyuan;PEI Zhongcai(School of Automation Science and Electrical Engineering,Beihang University,Beijing 100191,China)

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191

出　　处：《北京航空航天大学学报》2023年第7期1563-1572,共10页Journal of Beijing University of Aeronautics and Astronautics

摘　　要：针对非线性、强耦合并带有不确定性干扰的四旋翼无人机模型,提出了一种改进粒子群算法-径向基(PSO-RBF)神经网络自适应滑模控制器。在对RBF神经网络自适应滑模控制器进行控制量平滑改进的基础上,利用改进的具有全局寻优能力的PSO算法来调整RBF神经网络的拟合参数,从而进一步提升网络的拟合能力。根据实际四旋翼的模型参数,搭建四旋翼的动力学模型,通过Lyapunov理论验证了系统的稳定性。仿真结果表明:与RBF神经网络自适应滑模控制器和双闭环PID控制器相比,改进PSO-RBF神经网络自适应滑模控制器可以在一个控制周期内寻找到合适的控制量,其调节时间分别提升约50%和75%;改进PSO-RBF神经网络自适应滑模控制器具有轨迹跟踪速度快且准、抗干扰能力强和鲁棒性好的特点。An improved particle swarm optimization-radial basis function(PSO-RBF)neural network adaptive sliding mode controller is proposed for quadrotor systems with nonlinearity,strong coupling,and inaccurate interference.First,based on smooth improvement of the control amount of the RBF neural network sliding mode controller,an improved particle swarm optimization with global optimization capability was used to adjust the fitting parameters of the RBF neural network,thus improving the fitting ability of the network.Next,a dynamic model of quadrotor was built according to themodel parameters of actual quadrotors,the stability of which was then proved by Lyapunov theory.In contrast to the RBF neural network adaptive sliding mode controller and the double closed-loop PID controller,the improved PSO-RBF neural network adaptive sliding mode controller can find the appropriate control quantity in one control cycle,and its adjustment time is about 50%and 75%faster than that of RBF neural network adaptive sliding mode controller and double closed-loop PID controller,respectively.The simulation results show that the improved PSO-RBF neural network adaptive sliding mode controller featuresfasttrack speed with high accuracy,strong disturbance rejection and better robustness.

关 键 词：四旋翼无人机 粒子群算法 径向基神经网络 自适应滑模控制 轨迹跟踪 抗干扰 

分 类 号：V249.122[航空宇航科学与技术—飞行器设计]