Unmanned Aerial Vehicles parameter estimation using Artificial Neural Networks and Iterative Bi-Section Shooting method

• A discrete time model for the “OS4” Quadrotor is developed.
• Three parameter estimation methods namely, Iterative Bi-Section Shooting (IBSS), Artificial Neural Network (ANN), and the novel new Hybrid ANN–IBSS, were proposed.
• Performance of presented methods was evaluated under two scenarios; one using noise free trajectories and another with injected noise trajectories.
• Hybrid ANN_IBSS estimation results are the most accurate and fastest among all presented methods.

Quadrotor Unmanned Aerial Vehicles (UAVs) can perform numerous tasks fearless of unnecessary loss of human life. Lately, to enhance UAV control performance, system identification and states estimation has been an active field of research. This work presents a simulation study that investigates unknown dynamics model parameters estimation of a Quadrotor UAV under presence of noisy feedback signals. The latter constitute a challenge for UAV control performance especially with the presence of uncertainties. Therefore, estimation techniques are usually used to reduce the effect of such uncertainties. In this paper, three estimation methods are presented to estimate unknown parameters of the “OS4” Quadrotor. Those methods are Iterative Bi-Section Shooting method “IBSS”, Artificial Neural Network method “ANN”, and “Hybrid ANN_IBSS”, which is a novel method that integrates ANN with IBSS. The “Hybrid ANN_IBSS” is the main contribution of this work.Percentage error of the estimated parameters is used to evaluate accuracy of the aforementioned methods. Results show that IBSS and ANN are capable of estimating most of the parameters even with the presence of noisy feedback signals. However, their performance lacks accuracy when estimating small-value parameters. On the other hand, Hybrid ANN_IBSS achieved higher estimation accuracy compared to the other two methods. Accurate parameter estimation is expected to enhance reliability of the “OS4” dynamics model and hence improve control quality.