Research On System Identification Of Small-scale Unmanned Rotorcraft Base On Subspace Method

Small-scale unmanned rotorcraft is widely used both in military and civilian areas with the characteristics of light weight, small size, vertical taking off and landing. In order to achieve the precise control of the rotorcraft, an accurate model is strongly required. However the complex mechanical properties of the system make the theoretical modeling particularly difficult, now the common method being used is dynamics modeling and system identification. Subspace identification method has been developed rapidly in recent years. It aims at the identification for the state-space modeling and can be applied in multi-input and multi-output system.This paper focuses on applying subspace identification methods to determine the attitude channel modeling of small-scale rotorcraft which is in hover state. The details as follows:1. Describes the basic theory of subspace identification under an unified framework, gives the specific steps of calculation, contrasts the classic subspace identification methods and analyzes the advantages and weaknesses by simulation.2. Use closed-loop subspace identification method to overcome the unstable characteristic of small-scale rotorcraft system. In order to make convenience for identify rotorcraft system, this paper combines traditional two steps close-loop with orthogonal projection theory and gives the subspace identification method based on twice orthogonal decomposition, thus ensure the accuracy and reduce the computation. The simulation result proves the effectiveness of the algorithm.3. Form the basic dynamics equation of unmanned rotorcraft established the full state amount system model, completed the linearization under small perturbation assumption and divided into three attitude angle channel state subspace equations, gives the idea and steps for close-loop subspace identification method.4. Designed experiments based on orthogonal projection method, achieved subspace identification method in attitude channel model and verified with cross-validation method, the simulation results ensure the correctness, meanwhile compares the results of different orders to illustrate the realness of the system model and gives the criteria of order selection by judging the outputs variance. The simulation results proved the feasibility of subspace identification algorithm.