High-precision positioning system based on residual modeling and GNSS integrated inertial navigation system

Abstract

The invention discloses a high-precision positioning system based on residual modeling and a GNSS integrated inertial navigation system. The system mainly comprises a preprocessing module, a point cloud registration module and a repositioning module. The method is mainly based on residual modeling, the accuracy and robustness of a matching algorithm are improved by modeling residual distribution into a mixed probability model, hidden variables, residual distribution parameters and rigid transformation are decoupled through an expectation maximization algorithm, and efficient alternate optimization solving is achieved. Accurate lane-level map information in the structured urban environment is given. According to the method, a real-time, accurate and robust positioning result can be output in an open and high-dynamic urban scene, and the method can be widely applied to the fields of unmanned positioning navigation, advanced aided driving technologies and the like.

Description

technical field [0001] The invention belongs to the field of intelligent transportation and a high-precision positioning system, in particular to a high-precision positioning system based on a laser radar and GNSS combined inertial navigation system. Background technique [0002] High-precision positioning is the core module of the unmanned driving system. Through the high-precision positioning system, the environmental perception and motion decision-making module can read the passable area from the map, thereby verifying the output results and concentrating computing resources in the passable area, which greatly enhances the overall efficiency of unmanned driving. System security and real-time performance. The input of the high-precision positioning system is the current frame point cloud, GNSS integrated inertial navigation system and high-precision three-dimensional map, and the output is the rigid transformation of the current frame point cloud in the high-precision thr...

AI Technical Summary

Problems solved by technology

However, both the fixed threshold and the hyperparameters in the robust estimation need to be manually set, which cannot be adapted to complex and changeable traffic scenarios; residual modeling techniques usually use unimodal distributions such as Laplace distribution, and the modeling accuracy and The solution accuracy of rigid transformation cannot be guaranteed; and the above three types of algorithms do not consider IMU constraints in the solution process

To sum up, the current point cloud registration algorithm cannot be used as the core module of unmanned high-precision positioning, and it does not have the promotion in technology research and development and application.

Images