Multi-rotor UAVs have been increasingly used in precision tasks, driving research into novel designs that improve their controllability. One such design allows for decoupling thrust and moment directions, improving manoeuvrability. However, it introduces cross-interference effects between propellers, which reduce the thrust that the propeller can produce. The cross-interference is characterised by coefficients that can be identified offline by fitting a sufficiently large set of measurements via system identification methods. This work aims to address the problem by online estimation of the interference effect. Assuming the position of a mass attached to the UAV can be measured and the mass is subject to the forces generated by the propellers in the structure, the following research questions are formulated:
(i) Can the full state of the system, hereby including the interference weights, be reconstructed? (ii) What is the minimal measurement set required? and (iii) Can a functioning observer be built? How does the observer perform in simulation?\\
The observability was investigated with a constant input thrust assumption (linear, time-invariant system) and without it (linear, time-variant system). In both cases, it was shown that the full state is observable if both propellers are active, with the output measurement only including the position of mass $ m_{L} $ in 2D.
An observer was designed and tested through simulations for varying thrust directions and varying pole configurations. The results confirmed that the estimated states converge to the real states in each case.
A Lyapunov-based stability analysis verified that all tested observer configurations are asymptotically stable. These results demonstrate that the cross-influence effects can be estimated in real time, enabling future control strategies to include online correction terms without offline system identification.