Skip to Main Content
An approximate particle filter and sensor fusion solution in pedestrian tracking for integrating a global position system (GPS) and dead reckoning (DR) from a Stochastic Bayesian perspective for State Space Models is proposed. It is assumed that estimates obtained from the GPS receiver are correct if the GPS quality is good. Therefore estimates from the GPS receiver serves as a primary input into the integrated system whenever both the GPS receiver is on and the signal quality is good. The Suboptimal estimation technique for the DR sensor is based on particle filtering. The DR module takes inputs from an accelerometer based pedometer and a particle filtering method is used for estimating the state of the DR sensor. A sensor fusion model incorporating an exponential smoothing-based smoother/filter/predictor is finally used to integrate the DR sensor and the GPS receiver. The output of the sensor fusion algorithm is constructed as a function of the GPS receiver and DR sensor. At any state, higher weight is assigned to the sensor with minimum error. GPS positioning is weighted more heavily as long as the GPS parameters (DOP, number of satellites, signal quality) indicates good and reliable performance.