System Identification and Modeling of a Multirotor UAV: A Comparative Study

System Identification (SI) has been applied immensely in control and simulation applications. There are multiple types of identification methods, each of which addresses different issues and is suitable for certain types of problems. These methods can be categorized into three main categories: white box, grey box, and black box. Moreover, each category can be further divided into methods for identification of linear and nonlinear models. Since there are so many SI techniques available in each section, ranging from traditional frequency-response and state-space system identification to neural-network-based Auto Regressive Exogenous Input (ARX) or its nonlinear variant (NARX) and deep-learning-based methods such as Long Short-Term Memory (LSTM) and Gradient Boosting. It takes a lot of expertise and a careful survey of the various options to choose the correct modeling strategy. This paper surveys different approaches to system identification for multirotor UAVs, discusses promising methods and more importantly, implements them for a quadcopter multirotor UAV test platform. Finally presents the experimental results comparing the accuracy and predictive capability of each method.