Real-Time System Identification Using Deep Learning for Linear Processes With Application to Unmanned Aerial Vehicles

Abdulla Ayyad,Mohamad Chehadeh,Mohammad I. Awad,Yahya Zweiri

doi:10.1109/access.2020.3006277

Abstract

This paper proposes a novel parametric identification approach for linear systems using Deep Learning (DL) and the Modified Relay Feedback Test (MRFT). The proposed methodology utilizes MRFT to reveal distinguishing frequencies about an unknown process; which are then passed to a trained DL model to identify the underlying process parameters. The presented approach guarantees stability and performance in the identification and control phases respectively, and requires few seconds of observation data to infer the dynamic system parameters. Quadrotor Unmanned Aerial Vehicle (UAV) attitude and altitude dynamics were used in simulation and experimentation to verify the presented methodology. Results show the effectiveness and real-time capabilities of the proposed approach, which outperforms the conventional Prediction Error Method in terms of accuracy, robustness to biases, computational efficiency and data requirements.

Highlights

Since the third industrial revolution, system identification has been a key element in the development of autonomous technologies in a wide set of industrial applications
We show that this identification method can be performed in real-time such that a Unmanned Aerial Vehicle (UAV) adapts to changes to its own physical dynamics during a flight mission
Simulation results are compared against prediction error methods (PEM) as a well-established system identification method and the non-parametric tuning rules as an optimal controller design criteria

Summary

Introduction

Since the third industrial revolution, system identification has been a key element in the development of autonomous technologies in a wide set of industrial applications. To meet the aforementioned requirements of autonomy, extensive research has been carried out to develop effective methods of system identification and adaptation These methods are generally classified as parametric or non-parametric depending on the control requirements and design constraints. Several studies in the literature applied these techniques to UAV operation with accurate identification results [18]–[23] These methods require extensive data generation and accurate selection of optimizer initial conditions, which demand human experience and cause susceptibility to data biases and overfitting. Most of these methods are computationally expensive and not suitable for real-time applications; they are instead applied offline to process an abundance of previously collected flight or operational data [18]–[22]

Objectives

Methods

Results

Conclusion