Embedded Tightly Coupled INS/DGPS-DGAL Navigation Filter on a Mass-Market Single-Board Computer

摘要

This work contributes to the highly relevant research topic of integrating sophisticated navigation algorithms on mass-market embedded platforms for use, for example, on ship pilotage information systems that assist pilots in harbor maneuvers and lock entries. Therefore, the integration of a tightly coupled navigation filter on a Raspberry Pi 4B mass-market single-board computer is presented. An existing unscented Kalman filter (UKF) is integrated into a specially adapted multi-threading framework which uses classic real-time scheduling techniques. The UKF fuses measurements from an industrial-grade inertial measurement unit (IMU) and observations from global satellite navigation systems (GNSS). Hence, a full 6-DOF state estimation is achieved. The results represent a breakthrough for mobile and compact navigation for mass-market applications. To achieve high positioning performance and to enhance the number of usable satellites, recent available differential corrections of a virtual reference station (VRS) for the observations of the new Galileo (GAL) system are used in addition to GPS observations. Real word tests prove the online capability of the approach. Even with the additional GAL satellites, the computational load remains remarkable low. With an average total execution time of 0.45 ms it is proved that a tighly coupled UKF can be integrated on a Raspberry Pi 4B. The UKF achieves 1 m positioning accuracy which meets the International Maritime Organization (IMO) requirements in port areas; under GNSS signal degrading conditions, the average 2D error is about 1.5 m.