Precise Vehicle Localization based on Graph Optimization with Time-Relative RTK–GNSS

Abstract

The global navigation satellite system (GNSS) has recently been used for the autonomous control of unmanned vehicles. However, GNSS data is not utilized to its full potential for autonomously navigating vehicles in urban environments. This is largely owing to the possibility of the degradation of GNSS observables (e.g., multipath and poor satellite geometry) in an urban environment. The GNSS Doppler frequency can be used to derive an accurate velocity without the requirement for a base station. This velocity is quite tolerant to the strong multipath condition as compared with the GNSS pseudorange-based position solution. Thus, the Doppler velocity can be used to estimate vehicle trajectories. However, positioning errors accumulate over time owing to the error in the estimated Doppler velocity. To overcome the aforementioned issues, we leverage the advances made within the robotics community surrounding the pose-graph-based optimization and localization technique. The main idea of this study is that a loop closure of the pose graph is generated from a time-relative real-time kinematic GNSS (TR-RTK–GNSS) technique. The TR-RTK-GNSS is based on time-differential carrier phase positioning, which is a method comprising the implementation of a precise carrier-phase-based differential GNSS with a single low-cost GNSS receiver. As compared with the conventional RTK–GNSSs, we can directly compute the vehicle relative position using only a stand-alone GNSS receiver. The initial pose graph is generated from the accumulated velocity computed using a GNSS Doppler measurement. To cancel the accumulated error of the velocity, we use the TR-RTK–GNSS as the loop closure in the graph-based optimization frameworks. To confirm the effectiveness of the proposed technique, two kinematic positioning tests were performed using an unmanned aerial vehicle and a ground vehicle. In conclusion, we can estimate the vehicle trajectory with centimeter accuracy in an open-sky environment and with a few tens of centimeters accuracy in an urban environment using only a stand-alone GNSS receiver. Based on the test results, we concluded that the proposed technique is effective for estimating the precise vehicle trajectory.