基于神经网络的无人机姿态自适应控制仿真

针对四旋翼无人机姿态控制中模型不完整、部分参数和扰动不确定的问题,提出了一种基于神经网络的自适应控制方法,采用RBF神经网络对无人机姿态动力学模型中不确定和扰动部分进行学习,设计了以类反步法为基础,包含反馈控制和神经网络控制的自适应控制器,实现了对未知动态的准确逼近,解决了传统控制方法中过于依赖精确模型的问题。同时设计了神经网络的权值自适应律,实现了控制过程中的在线学习和调整,并且通过李雅普诺夫方法证明了闭环系统的稳定性。仿真结果表明,在存在较大扰动的情况下,上述控制器可得到很好的控制效果,可以实现误差的快速收敛,具有较好的鲁棒性和自适应性。

Simulation Study on Adaptive Attitude Control for a Quadrotor Based on Neural Network

In this paper,an adaptive attitude control scheme for quadrotor is proposed based on neural network,dealing with the attitude control problem including unmodeled uncertainties,parametric uncertainties and external disturbances.The uncertainties and disturbances in attitude dynamic model were learnt based on RBF neural network.The designed adaptive controller was based on backstepping method,including a feedback controller and a neural controller,which approximates unknown dynamics exactly and sloves the accurate model-independency problem in traditional control scheme.Meanwhile,the adaptive law for the weight of neural network was given to achieve the online learning and adjusting in the control process.The close-loop stability was proven with Lyapunov function.Simulation results are given to validate the effectiveness of the designed controller under the extreme disturbances,with rapid rate of convergence,showing better robustness and adaptivity.