Array PPP-RTK: A High Precision Pose Estimation Method for Outdoor Scenarios

Abstract

Advanced driver-assistance system (ADAS) and high levels of autonomy for vehicular applications require reliable and high precision pose information for their functioning. Pose estimation comprises solving the localization and orientation problem for a rigid body in a three-dimensional space. In outdoor scenarios, the fusion of Global Navigation Satellite Systems (GNSS) and inertial data in high-end receivers constitutes the baseline for ground truth localization solutions, such as Real-Time Kinematic (RTK) or Precise Point Positioning (PPP). These techniques present two main disadvantages, namely the inability to provide absolute orientation information and the lack of observations redundancy in urban scenarios. This paper presents Array PPP-RTK, a recursive three-dimensional pose estimation technique which fuses inertial and multi-antenna GNSS measurements to provide centimeters and sub-degree precision for positioning and attitude estimates, respectively. The core filter is based on adapting the well-known Extended Kalman Filter (EKF), such that it deals with parameters belonging to the SO(3) and GNSS integer ambiguity groups. The Array PPP-RTK observation model is also introduced, based on the combination of carrier phase measurements over multiple antennas along with State Space Representation (SSR) GNSS corrections. The performance assessment is based on the real data collected on an inland waterway scenario. The results demonstrate that a high precision solution is available 99.5% of the time, with a horizontal precision of around 6 cm and heading precision of 0.9 degrees. Despite the satellite occlusion after bridge passing, it is shown that Array PPP-RTK recovers high accurate estimates in less than ten seconds.