System Identification Method Research Of Small-sized Unmanned Helicopter

The small-sized unmanned helicopter owns the characteristics of small size, low cost, norequirement of launch systems, vertical taking-off and landing, hovering, high flexibility, andgood flying qualities. So in a number of specific environments, it has become an ideal carryingplatform of application systems. In order to control the unmanned helicopter accurately andreliably, it is necessary to have a precise mathematical model of its dynamics. Systemidentification is the most effective method to obtain mathematical model of the small-sizedunmanned helicopter.In this thesis, system identification method of small-sized unmanned helicopter is studiedin detail, and results with practical value are achieved. Main works of the thesis are as followingthree aspects:1) An effective and simple identification method was designed. The method is calledimproved simulated annealing algorithm combined with PEM (Prediction Error Method).Firstly, based on characteristics of model identification of the small-sized helicopter, thecriterion function and two basic requirements are stated. The improved simulated annealingalgorithm combined with PEM and two other identification methods are proposed in themethodology part. Next, these three identification methods are compared through simulationexperiments, and the improved simulated annealing algorithm combined with PEM has beensieved out initially. After that, according to real flight data, the effectiveness of theidentification method of small-sized helicopter under complex conditions is further analyzed.Finally, the effectiveness of the identification method is validated by both the simulation resultsand the real flight experimental results.2) A set of easy-operating environmental tools for identification process was developed.System identification is a kind of work that needs good planning and engineering. Sinceexperimental environments are complex and parameters to be identified for small-sizedunmanned helicopter are too many, the thesis organizes all the important sections of identification process in order to simplify future operations. These sections include: experimentdesign and data collection, data pre-processing, estimation of initial parameter values, Morrissensitivity analysis of helicopter model, application and validation of the identification method.The environmental tools for identification process have been used successfully during theexperimental stages.3) An efficient distributed computing platform has been established. The improved simulatedalgorithm can run concurrently. So in order to take the advantages of identification algorithmand MATLAB math capabilities, a MATLAB distributed computing cluster was built.Experiments show this computing platform is convenient for algorithm implementation, easyto deploy, and has good performance.