Parameter Estimation Techniques for Helicopter Dynamic Modelling

Abstract

Abstract-This paper investigates the problem of estimating parameters in a state-space helicopter dynamical model for helicopter controller design and evaluation. Two methods, PEM (the prediction error method) and RBPF (the RaoBlackwellized particle filter-based method), are presented in the paper. Computer simulations are carried out to assess and compare the performance of the two methods with different assumptions on the initial conditions of the parameters to be estimated. Based on this assessment, an estimation scheme that combines the two methods is proposed. The proposed scheme is then applied to real data from EH101 helicopter flight tests and a vertical model including rotor coning dynamics is identified.

A Two-Stage Method for the Parametric Identification of Scale-Model Helicopter Dynamics

“…In the framework of the RBPF presented in Li's papers, 19,20 the EKF is used to update the states and perform the measurement update. Since the UKF can handle the non-linearity better than the EKF, 15 it is interesting to test the performance of a RBPF that incorporates a UKF, i.e.…”

Longitudinal Aerodynamic Parameter Estimation Using Neural Network and Gauss-Newton Method

“…There are several approaches to model building. The so called "white box" approach starts from the first principle and a model is derived from basic physical laws that govern the behavior of the system [5]. This approach works for relatively simple systems, but its complexity increases manifold for recently introduced highly augmented, high performance aircraft.…”

“…This approach works for relatively simple systems, but its complexity increases manifold for recently introduced highly augmented, high performance aircraft. On the other hand, the "black box" approach generates a model based entirely on the input/output measurements of the system without trying to model the internal physical mechanism of the system [5]. Since we wish to estimate and validate such models from measured input-output data, the Artificial Neural Networks (ANNs) provide an alternative approach to model building.…”

“…The Cramer-Rao bounds [1,4] are also obtained which are measures of confidence in each parameter estimate. Strictly speaking, there is a lower bound on the variance of each model parameter, but they are commonly used as a measure of the relative quality of identification rather than an absolute measure [5]. The Delta and Modified Delta methods did not have the provision to estimate Cramer-Rao bounds along with the estimates.…”

Estimation of the Uncertainty in a Helicopter Dynamic Model Identified by the Subspace-based Method Using Bootstrap Techniques

“…The problem of helicopter dynamic modelling via system identification has been studied for many years and a large number of papers have been published over the last two decades (see e.g. Tischler and Remple (2006), Li et al (2007a) and references therein). Of great importance, in particular for robust controller design, is knowledge of the uncertainty in the identified model.…”