Multirotors from Takeoff to Real-Time Full Identification Using the Modified Relay Feedback Test and Deep Neural Networks

Abstract

Low cost real-time identification of multirotor unmanned aerial vehicle (UAV) dynamics is an active area of research supported by the surge in demand and emerging application domains. Such real-time identification capabilities shorten development time and cost, making UAVs' technology more accessible, and enable a variety of advanced applications. In this paper, we present a novel comprehensive approach, called DNN-MRFT, for real-time identification and tuning of multirotor UAVs using the Modified Relay Feedback Test (MRFT) and Deep Neural Networks (DNN). The first contribution is the development of a generalized framework for the application of DNN-MRFT to higher-order systems. The second contribution is a method for the exact estimation of identified process gain which mitigates the inaccuracies introduced due to the use of the describing function method in approximating the response of Lure's systems. The third contribution is a generalized controller based on DNN-MRFT that takes-off a UAV with unknown dynamics and identifies the inner loops dynamics in-flight. Using the developed generalized framework, DNN-MRFT is sequentially applied to the outer translational loops of the UAV utilizing in-flight results obtained for the inner attitude loops. DNN-MRFT takes on average 15 seconds to get the full knowledge of multirotor UAV dynamics and was tested on multiple designs and sizes. The identification accuracy of DNN-MRFT is demonstrated by the ability of a UAV to pass through a vertical window without any further tuning, calibration, or feedforward terms. Such demonstrated accuracy, speed, and robustness of identification pushes the limits of state-of-the-art in real-time identification of UAVs.

Real-Time Adaptive Dynamics Based State Estimation Scheme for Unmanned Aircrafts

“…The model structures used for identification were proposed by [28] and confirmed experimentally in [22], [29], [31]. The vertical motion dynamics (i.e.…”

Dynamic Based Estimator for UAVs with Real-time Identification Using DNN and the Modified Relay Feedback Test

“…This work further develops the controller tuning approach presented in [18] for state estimation. We showed that the SISO models used for channel wise parameter identification is able to capture the dynamics of the UAV presented in literature.…”

“…Note that we have omitted the yaw moment as estimation of the yaw dynamics is not of interest in this work, where the focus is devoted to fast transient dynamics that are present in aggressive maneuvers. Moreover, control and estimation for yaw dynamics is simpler due to the presence of full state measurements and the lower relative degree of the system [18]. Therefore in this work we are using a kinematic estimator for the yaw states based on [1].…”

“…( 14). The time delay in position measurement e τx,ys and the equivalent gain K x,y are not used in the estimators, but they have essential role in the design of optimal feedback controllers [18]. Each of the translational drag time constants is used to form the drag matrix T Λ .…”

“…The number of networks in the second set depends on the number of classes in the first set. This was imposed by the composition approach described in [18]. In this work, the first set is used to estimate the SOIPTD processes parameters in Eqs ( 15) and ( 17), and the second set is used for the FOIPTD parameters shown in Eq.…”