Initialization of a Pedestrian Navigation System Using a Transfer Alignment Approach

Abstract

Pedestrian navigation techniques are used to provide positioning information in lieu of or to supplement more traditional systems such as GNSS. A challenge facing users of pedestrian navigation algorithms is the acquisition of initialization data. Existing initialization methods for pedestrian navigation methods assume the availability of GNSS position and velocity and of standstill periods that can be used to provide attitude information. This thesis presents a method that performs the attitude initialization for a pedestrian navigation system (PNS) while the pedestrian is riding inside a vehicle. An algorithm is presented which uses data from a vehicle-mounted IMU and a pedestrian-mounted IMU to detect when the two sensors form a rigid body. Once rigidity has been established, a solution to Wahba’s Problem is used to solve for the misalignment between the two systems. The misalignment is used in conjunction with the vehicle’s attitude to initialize the PNS before the pedestrian leaves the vehicle. Performance analyses of the rigidity detector and misalignment estimation are shown using both simulated and real data, while a performance analysis for the PNS initialization procedure is shown using real data. The method is implemented using IMU data gathered using a Vectornav VN-300 GPS/INS and a Lincoln MKZ instrumented with a KVH 1750 IMU. Results show that the proposed method can initialize a PNS with a mean initial heading error of 7.3 degrees, with the error being attributable to violations of the rigidity condition.