Data-Driven Modeling for Multirotor Autonomous Control

Abstract

The precise trajectory control of UAVs, including multirotors, is challenging due to uncertainties such as modeling errors, unmodeled aerodynamic forces, and wind disturbances. To address this issue, this paper focuses on real-time modeling of quadrotor motion exploiting recently developed data-driven methods. By constructing a dynamical model in real-time using data collected during flight, it is expected to develop a dynamical model that incorporates uncertainties arising from modeling errors and disturbances. In particular, the data-driven modeling method called Real-Time Update Dynamic Mode Decomposition (RTDMD), an extension of the DMD algorithm with sequential update rules, is employed for the real-time modeling of dynamical models. Identification experiments using experimental data demonstrate that the RTDMD successfully obtained a dynamical model for estimating the quadrotor’s state.

Publication

T. Shiotsuka, M. Bando, S. Hokamoto, “Data-Driven Modeling for Multirotor Autonomous Control” in AIAA SCITECH 2024 Forum, 0568, 2024. [Paper]