Rapid Position Initialization for Automated Automotive Applications

Abstract

Research work in autonomy and specifically automotive autonomy has been progressing swiftly over the past decade. A key element for successful autonomous applications is reliable, accurate and precise position information. Position is necessary to initialize perception sensors such as a Lidar or a camera and to support lane navigation. Global Navigation Satellite Systems (GNSS) and GNSS aiding sensors play a key role in providing precise and reliable initial position. There has not been any detailed study published on achieving rapid sub-metre-level accuracy using low-cost mass market sensors. In this research, GNSS and Inertial Measurement Unit (IMU) sensors along with software constraints are reviewed to initialize a navigation system with better than one metre accuracy in a few seconds to tens of seconds. To achieve this objective, the Precise Point Positioning (PPP) augmentation is applied on low-cost, multifrequency, multi-constellation GNSS measurements, external global ionospheric data are used, carrier-phase ambiguities are fixed to integers, and GNSS measurements are tightly coupled with a Microelectromechanical system (MEMS) IMU. The algorithm was examined in open sky, suburban and urban environments in static and kinematic datasets. The results show that rapid sub-metre positioning is possible in different environments, although not as easily achievable in the static urban case. Results show that, on average, regardless of the scenario, while conventional single-point positioning achieves accuracies of 7 to 8 metres in less than 30 seconds, the enhanced solution produces an accuracy of 8 to 9 dm. Improvements are more noticeable in the suburban and urban environments where the solution is more resilient to poor satellite geometry and multipath compared to single-point positioning. These impressive results clearly show that low-cost hardware equipment with GNSS-PPP augmentation, a MEMS IMU and dynamical constraining can be used for quick position initialization in autonomous applications.