AI-driven kinematics and experimental control of 3-SRR mechanism for UAV stabilization

This paper investigates the integration of artificial neural networks (ANN) to the inverse kinematics and the control of parallel manipulators. The method is applied to a 3-SRR parallel manipulator for applications such as the stabilization of unmanned aerial vehicle (UAV). The study concludes with experimental validation filling the gap between theoretical modelling and practical implementation. The kinematic behavior of the manipulator is simulated in Simscape, based on real-time PID control with the implementation of ANN. The manipulator was able to stabilize roll and pitch movements under dynamic conditions, which are modeled, tested, and experimentally validated. Using extensive data gathered from Simscape simulations, an ANN predicts the desired joint angles, successfully solving the non-linear equations associated with inverse kinematics. This approach, relying on ANN, offers a more efficient and accurate solution compared to traditional methods, which significantly enhances the manipulator’s adaptability and control. An experimental setup is designed to replicate the manipulator’s performance in UAV applications. It integrates real-time sensor feedback, PID-controlled actuators, and ANN to achieve dynamic stabilization. Results show an excellent agreement between theoretical predictions and experimental outcomes, where the manipulator consistently achieves precise stabilization, particularly under extreme roll and pitch angles. Key measurements, such joint angle alignment accuracy and validation using the IMU sensor, demonstrate the robustness of the proposed methodology.