Skip to content

NN-Based Fuzzy Adaptive Control for UAVs Using Cascaded Architecture Under Unmatched Uncertainties and Interference

2026 · IEEE Transactions on Automation Science and Engineering · Vol 23, pp. 19331-19342 · 0 citations · 29 references

Abstract

This paper proposes a fuzzy-logic-system-based adaptive control (FLSAC) strategy for position and attitude tracking of a quadrotor subject to lumped nonlinear uncertainties and bounded external disturbances. The controller adopts a cascaded structure, in which the outer position loop and inner attitude loop employ channel-wise time-varying filtered-error gains generated online by a fuzzy self-learning-gain module. An RBFNN-based compensation mechanism with fixed Gaussian basis functions and a scalar adaptive parameter in each control loop is introduced to compensate for unknown nonlinear dynamics and disturbances without adapting the complete ideal weight matrix. Adaptive control and parameter-update laws are derived using Lyapunov analysis, and the uniform ultimate boundedness of the position- and attitude-tracking errors is established, with the ultimate bounds given explicitly in closed form. Comparative MATLAB/Simulink simulations are conducted under nominal conditions, external disturbances, and payload-induced parameter mismatch, using proportional–integral–derivative control, sliding-mode control, and RBFNN-based adaptive control as baselines. Axis-wise and aggregate root-mean-square error and integral absolute error metrics show that the proposed FLSAC consistently achieves the lowest aggregate position and attitude RMSE under the three evaluated conditions, while a performance trade-off remains in the yaw channel. The results demonstrate that FLSAC provides accurate and robust trajectory tracking under external disturbances and model-parameter mismatch while keeping the online self-learning gains bounded. Note to Practitioners—Quadrotors used in transportation, inspection, and monitoring often carry payloads of different weights. Payload changes, modeling errors, and external disturbances can reduce tracking accuracy, while conventional controllers may require repeated retuning. This paper presents a control strategy that automatically adjusts its response according to the tracking error and uses a lightweight online learning mechanism to compensate for unmodeled effects and disturbances. Only one value is updated online in each control loop, reducing computational and tuning burdens for onboard implementation. Simulations show that the same controller settings maintain stable and accurate tracking under nominal operation, composite disturbances, and a fixed additional payload without manual retuning. This may reduce field commissioning effort for tasks with changing payload requirements. Hardware-in-the-loop and flight tests are still needed to evaluate sensor noise, actuator dynamics, and onboard computing limitations.

View source

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.