四旋翼PIDNN控制器设计

Design of PIDNN Controller for Quadrotor Aircrafts

针对传统四旋翼PID控制器参数整定困难和控制效果难以达到最优的问题,综合了传统PID控制器工程意义明确、参数整定简便以及神经网络的非线性映射和自学习的优点,构造了四旋翼神经网络PID(PIDNN)控制器。以神经网络的非线性映射特点和自学习能力优化了传统PID控制器的控制效果,借助PID控制器的构造特点,解决了神经网络层数、节点数和连接权重初值选取困难的问题。最后,通过仿真实验验证了算法的合理性和有效性。

In traditional quadrotor PID controller, parameters tuning is difficult and it is also difficult to achieve optimum control effect. To solve the problems, we constructed a quadrotor PID neural network （PIDNN） controller, which integrated the advantages of the traditional PID controller of clear engineering meaning and simple parameter tuning, with the advantages of neural network （NN） of nonlinear mapping and self-learning capability. The nonlinear mapping and self-learning capability of NN were used to optimize the control effect of traditional PID controller. By constructing the PID controller, the initial values of number of neural network layers, nodes and connection weights were determined. Finally, the simulation experiments verified the rationality and validity of the algorithm.