Research on Neural Network PID Control Algorithm for a Quadrotor
Weinan Gao
1, a
, Jialu Fan
1, b
* and Yannong Li
1, c
1
State Key Laboratory of Synthetical Automation for Process Industries, Northeastern University,
Shenyang 110819, China
b
e-mail: fanjialu@gmail.com
Keywords: Quadrotor;Parameter uncertainties; Neural network; PID controller
Abstract. Quadrotor is a kind of popular unmanned aerial vehicle which obtains prime advantages
in simple structure, vertically taking off and landing and hovering ability; hence it possesses wide
application prospects in reconnaissance and rescue, geological exploration and video surveillance.
However, attitude and position control of the quadrotor are challenging tasks because it is an under-
actuated system with strong nonlinear, coupling and model uncertainty characteristics. In this paper,
the dynamics model and the state space function of the micro-quadrotor are firstly established.
Then, a cascade control scheme is proposed to decouple the control system and a multivariate
RBF(Radial Basis Function) neural network control PID algorithm is proposed to realize robust
control of the quadrotor. This algorithm is not only characterized by simple structure and easy
implementation, but also capable of self-adaption and online learning. Simulation results show that
the proposed control algorithm performs well in tracking and under disturbances and model
uncertainties.
Introduction
Autonomous flight technology of quadrotor has widely applied prospects in reconnaissance and
rescue, scientific data collection and geological exploration [1]. However, since the quadrotor is a
multi-variable, strong-coupling, nonlinear system with parameter uncertainty, general control
algorithm cannot get a good performance. [2, 3] use linear control theory to design a decouple
controller for a quadrotor, but the algorithm can stabilize the system in non-equilibrium points. A
slide-mode controller is designed in [4]. [5, 6] directly use backstepping algorithm to design
controller. It is generally known that backstepping algorithm relies on the accuracy of the model, the
practical control result is not satisfying.
In this paper, in order to enhance the robustness of the quadrotor control system, based on RBF
multi-variable neural network PID algorithm, the attitude and position controllers of quadrotor are
designed. Simulation result shows that the designed controller can track the predicted trajectory,
stand the wind disturbance and adjust the parameter uncertainty.
Model of a Quadrotor
When the elastic deformation and vibration of a quadrotor is ignored, the quadrotor can be seen as a
rigid body with 6 DOF (degree-of-freedom). And the lifting power is provided by four motors in the
quadrotor. Based on the dynamics knowledge, the rotation matrix and transition matrix [1] are
cos cos sin sin cos cos sin sin sin cos sin cos
cos sin sin sin sin cos cos cos sin sin sin cos
sin sin cos cos cos
1 sin tan cos tan
0 cos sin
0 sin / cos cos /cos
R
T
θ φ θ φ θ
φ θ φ θ
φ φ
φ θ φ θ
− +
= + −
−
= −
(1)
, where
are respectively roll, pitch and yaw angles. Considering the wind disturbance in the
three axes of the quadrotor, based on Newton-Euler equation, the final dynamics model is
Applied Mechanics and Materials Vols. 719-720 (2015) pp 346-351 Submitted: 10.09.2014
© (2015) Trans Tech Publications, Switzerland Accepted: 29.10.2014
doi:10.4028/www.scientific.net/AMM.719-720.346
All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of TTP,
www.ttp.net. (ID: 202.118.21.100-08/01/15,03:58:15)