Identification of a dynamical model and its parameter estimation is one of the fundamental problems in field of robotics and system dynamics modelling. For the general case of object motion with six degrees of freedom (6- DOF), such as in the case of Unmanned Aerial Vehicle (UAV), the key physical parameters are its mass and tensor of inertias. Even though UAV mass and its geometry/topology are easily obtainable, an identification of inertia tensor is difficult considering that is not measurable by static tests. This paper presents a simple and effective method for on-line estimation of a rigid-body inertia based on two-wire pendulum and integrated sensor system. Herein, the test subject (i.e. UAV) is suspended by two thin parallel wires in such a way to form a bifilar torsional pendulum about the vertical axis. Using on-board sensors from UAV flight controller (FC) unit, the pendulum oscillations are recorded and processed to obtain trend-free and noise-free signals used in the final inertia estimation phase based on inertia- related properties of pendulum oscillations. The proposed identification algorithm is verified experimentally for two typical cases of suspended objects related to UAV control box and full UAV configuration.
Read full abstract