Enhanced geometry control powered by AI for UAVS with a robotic arm for compensating for disturbances

Abstract Unmanned Aerial Vehicles (UAVs) equipped with robotic manipulators have emerged as a powerful paradigm for advanced aerial manipulation tasks, including infrastructure inspection, emergency response, and precise object handling in hazardous or inaccessible environments. Despite their potential, integrating robotic arms introduces significant challenges due to strong dynamic coupling, time-varying payloads, and external disturbances such as wind and aerodynamic turbulence, which can severely compromise flight stability and control performance. Conventional linear and nonlinear contrDistributed Control Architecture strategies often struggle to cope with these highly nonlinear dynamics, particularly during aggressive manipulator motions. This paper proposes an AI-enhanced geometric control framework for stabilising UAV–manipulator systems under high-manoeuvrability conditions. By operating directly on the nonlinear configuration space, geometric control provides robust attitude and position stabilization while avoiding singularities and ensuring global stability. To overcome the limited adaptability of purely geometric controllers, artificial intelligence algorithms are integrated to predict and compensate for manipulator-induced disturbances, optimize thrust distribution, and dynamically adjust BLDC motor currents in real time. The integration of robotic manipulators with Unmanned Aerial Vehicles (UAVs) offers a transformative approach to aerial manipulation in hazardous environments. However, the shifting center of mass and aerodynamic coupling during arm movement present significant stabilization challenges. This paper proposes a hybrid control framework that combines Geometric Control on the SO(3) manifold with a Long Short-Term Memory (LSTM) neural network. The LSTM is specifically designed to predict and compensate for non-linear disturbances and dynamic coupling effects in real-time. Experimental results demonstrate that this AI-enhanced geometric approach provides superior attitude tracking and disturbance rejection compared to traditional PID and sliding mode control, significantly reducing oscillation during complex maneuvers. A quadrotor UAV equipped with a Cartesian robotic arm employs the proposed framework for compensation of disturbances affecting both position and orientation. Simulation results obtained using a high-fidelity CoppeliaSim–MATLAB environment demonstrate the effectiveness of the proposed approach. The UAV maintains stable flight and accurate attitude regulation while carrying relatively high payloads and executing rapid robotic arm maneuvers.