Fault Diagnosis Of Mobile Robot Based On Variable Structure Multiple Model Unscented Kalman Filter

With fast development of science and technology, robot is put into use in many civilianand military fields, such as the mechanical arm in factory, the medical robot, the robotic fishfor hydrological detection, the unmanned aerial vehicle and unmanned submarine for military,etc. As the increase of working time and complexity and change environment, thedependability will reduce and then influence the work efficiency even lead to completelyparalyze caused by mechanical fault or system fault, which may lead to serious loss.Therefore, it can be seen that the research of fault diagnosis for mobile robot has veryimportant significance.The fault diagnosis for mobile robot of used in industrial, scientific and military field canbe divided into three types: the method based on signal processing, the method based onknowledge and the method based on analytical model. Compared with the other system, it iseasy to obtain the dynamical model of mobile robot and the mathematical model of velocityand angular velocity sensors of mobile robot, therefore, the fault diagnosis method is basedmainly in analytical model based. The traditional method based on analytical model is mainlyin interacting multiple model. The quantity of system mode and the model-set can be veryhuge as the combined fault is taken into consideration. The competition of independent modelcan lead to low estimation accuracy and low fault diagnosis accuracy rate when use thetraditional method is used. The traditional method for nonlinear system is extended kalmanfilter which need to calculate the jacobian matrix, which has a huge amount of calculation andalso contain the features of low calculation accuracy and complex calculate.In this paper, combining the variable structure multiple model algorithm with theunscented kalman filter not only can solve the model competition caused by the combinedfault which leads to a large number of fault models, but also can solve the problem of lowcalculation accuracy caused by traditional extended kalman filter used in nonlinearsystems. The simulation results show that this method effectively improves the response timeand the accuracy of fault diagnosis of mobile robot.