Analytical Modeling and Dynamic Analysis of a Fully-Actuated Hexarotor with a 2-Link Manipulator

Aerial manipulation extends the use of aerial robots beyond observation by adding a robotic arm. Its motion changes the mass distribution and couples the arm dynamics with the platform response. Therefore, modeling these effects is important for motion control. This work considers an Unmanned Aerial Manipulator (UAM) based on HR01, a fully actuated hexarotor available in the Room~F laboratory of the Department of Management and Engineering, University of Padova, in Vicenza. The platform is equipped with an OpenMANIPULATOR-X. Only the first two revolute joints move, while the remaining arm joints are kept fixed. For this system, the thesis aims to develop and validate an analytical model and assess the tracking benefits of updating the dynamic terms used by a model-based controller during system motion. The analytical model is derived through the Euler--Lagrange formulation and adapted to the geometry and mass distribution of the real UAM. It is implemented in MATLAB/Simulink and compared with a physics-based model built from multibody components in \textsf{RotorSuite}. Static tests compare the center-of-mass (CoM) position and diagonal inertia components across joint configurations. Open-loop tests then compare the coupled responses of the two models under identical actuator inputs. A sensitivity analysis also quantifies how variations in manipulator link lengths, link masses, and an end-effector payload affect the modeled center-of-mass position and inertia. A model-based geometric controller (GC) is then applied to each simulation model using the same reference trajectory and feedback gains. The closed-loop tests compare fixed and online-updated mass, Coriolis and centrifugal, and gravity terms in the controller's inverse dynamics. The fixed terms are evaluated at the nominal configuration at rest, whereas the online terms are recomputed from the current state. These updates concern the controller, while both simulation models retain their full dynamics. Tests with selected terms updated separately show that updating the mass matrix and gravity vector together improves joint tracking and reduces attitude deviations for the investigated trajectory. Additional tests assess compensation of the body torque caused by the center-of-mass offset and show that a consistent change of the torque reference point leaves the actuator commands unchanged. Finally, steady-hover measurements from the real UAM are compared with the simulated states and rotor speeds of both models for different fixed manipulator configurations. These tests assess the models under closed-loop steady-hover conditions. The real platform uses its onboard Pixhawk~4 flight controller, while the simulations use the GC, so the comparison does not provide an experimental validation of the GC. The models reproduce the main rotor-speed redistribution, although the measured speed level is higher and requires further characterization of the tested platform.