Adaptive Deep Neural Network Sliding Mode Control for UAVs

Chen Zhang,Jing Xu,Fanbiao Li,Yugang Niu
DOI: https://doi.org/10.1109/fasta61401.2024.10595247
2024-01-01
Abstract:This paper proposes an adaptive attitude maneuver control strategy based on deep neural network (DNN) for quadcopters to improve the attitude tracking performance and attenuate the effects of nonlinear and uncertainty. Obtaining an accurate dynamic model is extremely challenging, due to the disturbances of wind speed, air pressure, temperature, and other environmental factors. To solve this issue, we establish an affine function model to describe the black box model of quadcopters. The DNN trained over a large time scale is used to approximate the nonlinearity and uncertainty. Then an adaptive sliding mode control (SMC) is combined with the DNN to achieve accurate tracking of attitude trajectory. An update algorithm is proposed to ensure the real-time performance of the controller. The Lyapunov based stability analysis is used to prove the stability of the stability of the closed loop system. The experimental results show the good performance of proposed control strategy.
What problem does this paper attempt to address?