Fault Estimation of Unmanned Helicopter Servo Based on Random Forest Algorithm

Research of unmanned helicopter has become more challenging since the complexity of system structures is increasing. As one of the most important parts of an unmanned helicopter system, the servo needs to be estimated online to guarantee the real-time analysis and control. In this paper, a real-time fault estimation system of unmanned helicopter servo is presented which using the random forest algorithm and the ground station. The random forest algorithm, which based on statistical machine learning, used to process the input samples with high dimensional features and establish combined classifier models. The historical data onto the unmanned helicopter flight arc collected as the training set to train the algorithm. The power supply voltage, power supply current, feedback current and feedback stroke of servo are selected as the input sample features. The judgment value of servo's running state is obtained, and the fault estimation of servo is realized. Finally, an improved ground station can provide a variety of interfaces to ensure the online analysis and control of servo. The proposed system is checked during test, convincing results are presented. Experiment proves that the system can be estimated effectively with the servo's running state with high accuracy and no overfitting.