High-Precision Positioning (GNSS+RTK) is an enhanced technology that uses ground reference stations and differential technology to improve the positioning accuracy of Global Navigation Satellite Systems (such as GPS, BeiDou) from meter-level tocentimeter-level (1-10 cm) enhanced technology.
This technology was once a tool in the professional surveying field, and has now become a key technical cornerstone of autonomous driving and intelligent transportation systems, enabling vehicles to precisely perceive their position in the lane.
The accuracy of ordinary GNSS single-point positioning is around 3-10 meters; errors mainly stem from ionospheric delay, tropospheric delay, and satellite orbit errors generated when satellite signals traverse the atmosphere.
RTK (Real-Time Kinematic) solves this problem through the idea of 'differential':
Set up a reference station: Establish a reference station on the ground with known precise coordinates to continuously receive signals from all visible satellites.
Calculate error correction values: The reference station compares its known coordinates with the coordinates obtained via GNSS in real-time, calculates the combined impact of various errors in the current area, and generates 'differential correction values'.
Data link broadcast: The reference station broadcasts these correction values in real-time to nearby mobile stations (such as autonomous vehicles) through low-latency communication links like 4G/5G and radio.
Terminal differential processing: The mobile station receives the correction values from the reference station and applies them to its own satellite observation data, thereby canceling out most common errors, allowing positioning accuracy to reach centimeter-level.
Although RTK's theoretical accuracy is very attractive, in the real world, it is not all-powerful. The main challenges are:
Signal Occlusion and Multipath Effect: In areas with tall city buildings (i.e., 'urban canyons'), tunnels, underground parking lots, or dense tree shade, GNSS signals are easily occluded, reflected, or attenuated, causing mobile stations to fail to lock enough satellites or receive severely distorted signals, resulting in RTK calculation failure and a sharp drop in accuracy.
Dependency on Communication Link: RTK requires stable, low-latency (typically <100 ms) data links to transmit correction values. Once communication is interrupted, high-precision positioning becomes invalid.
Therefore, in autonomous driving systems, GNSS+RTK rarely works alone. It is usually deeply fused withInertial Measurement Units (IMU),Visual Odometry (VO),Laser Radar (LiDAR) and others to build a robust combined navigation system.
Complementarity: RTK is responsible for providing absolute centimeter-level coordinate 'anchors' when signals are good to correct the long-term drift of other sensors. IMU and Visual/Radar Odometry, however, are responsible for providing high-frequency, continuous relative positioning for tens of seconds when GNSS signals are briefly lost (e.g., entering a tunnel), maintaining smooth and stable positioning.
Engineering Practice: Industry research and practice show that through tight coupling fusion of RTK/INS/Visual, even in complex environments with poor signals, centimeter-level positioning accuracy can be achieved, and the reliability of positioning results can be effectively evaluated. For example, a scheme combining OBD wheel pulse data can control position error within3.5 meters.
In summary, GNSS+RTK is the cornerstone for autonomous driving to obtain 'lane-level' absolute positioning, but its implementation relies on ground infrastructure, and its performance is also challenging in real-world environments. Therefore,multi-sensor fusion is the fundamental solution to ensure autonomous driving positioning systems are both precise and reliable.
