ELM-Assisted Particle Filter for INS/UWB-Integrated Quadrotor Positioning

Abstract

At present, with the wide application of the quadrotor, accurate positioning has become an increasingly important problem and needs to be considered. In this study, the inertial navigation system (INS) and ultra-wideband (UWB) technology are used together to collect the flight data of the quadrotor, and the particle filter (PF) algorithm will be employed to fuse the data information of the two sensors for reducing the error and obtaining the movement path of the quadrotor in the three dimensional (3D) space. Meanwhile, for making PF work more accurate, the extreme learning machine (ELM) is adopted to map the equation of state in the filtering process to make position information more reliable. In addition, ELM can also establish a new signal through mapping when UWB’s signal is interrupted, so that the whole system can work normally. To verify the effectiveness of the proposed method, a real experiment was carried out. The experimental results show that the ELM-assisted PF strategy has a good effect on INS/UWB-integrated navigation quadrotor positioning. When UWB signals are normal, compared with the single PF, the ELM-assisted PF is able to improve positioning accuracy by about 18.44%. When UWB’s signal is interrupted, compared with the least square support vector machine- (LS-SVM-) assisted PF, the ELM-assisted PF could improve positioning accuracy by about 1.15 m. On the whole, the proposed design algorithm not only improves the positioning accuracy but also can predict UWB’s signal when it is interrupted and thus make the filter work normally.