Research on Multi-Sensor Data Fusion Positioning Network Cooperative Control Algorithm in Commercial Vehicle Autonomous Driving Scenarios

Engineering notes

Experimental results show that the algorithm can maintain sub-meter positioning accuracy even in extreme scenarios such as GNSS signal degradation and partial sensor failure, reducing positioning errors compared to traditional fusion algorithms, achieving faster network collaborative response speed, significantly improving the environmental adaptability and engineering practicability of the commercial vehicle autonomous driving positioning system, and providing theoretical support and technical solutions for the engineering implementation of high-precision positioning in commercial vehicle autonomous driving.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

High-precision and high-robustness positioning is the core prerequisite for the safe implementation of commercial vehicle autonomous driving. However, it faces key bottlenecks such as insufficient positioning accuracy of a single sensor in complex scenarios like urban canyons, tunnel obstructions, and multi-path interference, as well as data asynchrony and heterogeneity, unreasonable network resource allocation, and poor multi-node coordination. These issues severely restrict the reliability and real-time performance of the commercial vehicle autonomous driving system. To address these problems, this paper conducts research on the integrated algorithm of multi-sensor data fusion positioning and network cooperative control in commercial vehicle autonomous driving scenarios. Firstly, a data preprocessing and spatio-temporal alignment model for multiple heterogeneous sensors (GNSS, IMU, LiDAR, visual camera) is constructed, and a dynamic reliability assessment mechanism is introduced to solve the fusion problems of uneven observation errors and data asynchrony of different sensors; Secondly, a multi-sensor fusion positioning algorithm based on improved factor graph optimization is designed, integrating map matching and inter-frame motion constraints, to achieve dual improvements in positioning accuracy and robustness in complex scenarios, breaking through the limitations of traditional discrete-time fusion methods in asynchronous observation processing; Then, a network control architecture based on distributed cooperative strategies is proposed, combined with an adaptive resource allocation mechanism, to balance positioning accuracy, computational overhead, and communication delay, achieving global collaborative optimization of multi-sensor nodes; Finally, based on real commercial vehicle road test data and standard datasets, multi-scenario comparative experiments are conducted to verify the effectiveness of the proposed algorithm. Experimental results show that the algorithm can maintain sub-meter positioning accuracy even in extreme scenarios such as GNSS signal degradation and partial sensor failure, reducing positioning errors compared to traditional fusion algorithms, achieving faster network collaborative response speed, significantly improving the environmental adaptability and engineering practicability of the commercial vehicle autonomous driving positioning system, and providing theoretical support and technical solutions for the engineering implementation of high-precision positioning in commercial vehicle autonomous driving.

Links and sources

• Official / arXiv page

Comments