Thrust estimation of quadrotor UAV using adaptive observer

In this work, we apply an adaptive observer for estimating the thrust generated by the individual rotors of a quadrotor UAV with uncertain actuator parameters, using a linear dynamic model of the system. In most of the previous works, the implementation of quadrotor control is based on the assumption of a known static relationship between the voltage signal to the motors and the thrust generated by the actuator-propeller assembly. However, due to variations in this voltage-to-thrust relationship, this approach may lead to undesirable results in practice. Feedback of the quadrotor thrust in the control system can provide more robustness to the system. In the absence of accurate knowledge of the actuator parameters, we use an adaptive observer for thrust estimation in this work. An adaptive observer is an algorithm for joint estimation of states and unknown parameters of a dynamical system. We use the adaptive Luenberger observer using two different cost functions, and apply a genetic algorithm to select the design parameters of the observer. We use low- and higher-order models for the actuator with the adaptive observer, and compare their performance. The results show that the estimates from the adaptive observer used with a low-order actuator model match the thrust values obtained from measurements with good accuracy.