Data Fusion Method For Vehicle GPS/DR Integration

Vehicle location is the one of the important technology in intelligenttransportation system. Global position system named GPS is of superior long-termerror performance, but poor short-term accuracy, while dead reckoning systemnamed DR has good position precision in short-term but poor in long-term.GPS/DR integration provides position data with high precision, frequency andreliability.On the base of analyzing the multi-sensor characteristic of vehicle DR system,the mechanics layout formulae in local horizontal reference frame were deduced.The decentralized kalman filters were founded, with which the angle movementdata offered by electro-compass and rate gyro and the line movement data suppliedby accelerometers and vehicle odometer were fused respectively, and the estimateoptimization of vehicle location states were obtained. The data fusion methodwere simulate.For calculating the rational range of acceleration and making the vehicle modelto accord with the vehicle actual movement, basing on the current-statistics modelwhich is reasonable to maneuver target relatively and the kinetic characteristic ofvehicle, the fuzzy logic was brought forward to fuse the data of vehicle state. Theresult of qualitative analyzing and matlab simulations showed that the locationprecision and the tracking capability were improved by calculating theacceleration range through fuzzy logic. Furthermore, in order to solve thedivergence which is likely to emergence in kalman filter, the adaptive kalmanalgorithm based on fuzzy logic to avoid filter divergence was designed which relyon the convergence criterion of kalman filter, and simulation experimentationwere carried through.In order to enhance the tracking capability of maneuver target, the accelerationin target movement model must be considered, thereby, the adaptive kalman filteralgorithm were put forward in which the acceleration were estimated by adaptiveneural fuzzy inferential system named ANFIS , and the structure of neural networkwas constructed. The state of target maneuver were fused by ANFIS according tothe character data distilled. The matlab simulations manifested that the locationprecision and the tracking capability were improved by the algorithm.To improve the precision and continuity of GPS/DR more, the planar stateequation were established, in which the velocity state was added to as aobservation. In order to improve the location continuity when the GPS signal lost,the federal kalman filter fusion algorithm for GPS/DR and the GPS/DR fusionalgorithm model based on artificial neural network named ANN were put forward,and the simulation was performed.On the viewpoint of engineering application, the fusion framework and thealgorithm used for vehicle GPS/DR integration were studied in systemic, and theresults regard as the base of vehicle integration product.