基于车路信息交互的车辆卫星定位协同定权方法

为确保基于卫星导航系统的车辆定位性能满足特定交通应用需求，本文针对导航卫星观 测量权重分配对复杂动态运行环境的跟踪适配问题，建立基于车路信息交互的车辆卫星定位协 同定位增强总体框架；基于加权最小二乘定位解算模式，设计基于轻量级梯度提升机建模与多车 信息综合决策的协同定权方法，提出面向学习建模通道的导航卫星伪距残差轻量级梯度提升机 建模方案，面向定权计算通道设计了基于多车信息综合决策的导航卫星观测量权重决策策略。 实验结果表明：运用轻量级梯度提升机构建伪距残差模型，相较于支持向量机、随机森林以及仅 基于基础卫星观测特征的轻量级梯度提升机，均方根误差分别降低了62.1%，29.9%，60.4%；运用所建立的预测模型协同定权所得水平误差标准差，相对于等权和卫星仰角/信噪比组合定权策略分别降低了48.5%和47.6%。研究结果对于充分发挥车路协同系统模式下，信息交互机制的核心优势和优化车辆卫星定位性能具有重要意义。

Cooperative Weighting Method for Satellite-based Vehicle Positioning Based on Vehicle Infrastructure Information Interaction

In order to satisfy the performance requirement of vehicle positioning based on the Global Navigation Satellite System (GNSS), the tracking and mapping of the complicated and dynamic vehicle operation environment by the weight assignment to the observable GNSS satellites are concerned. An overall framework for the enhanced collaborative positioning based on information interaction is established. Based on the Weighted Least Squares (WLS) navigation calculation scheme, a cooperative weighting method is designed based on the Light Gradient Boosting Machine (LightGBM) modeling and comprehensive decision-making considering multiple neighboring vehicles. In this method, a modeling solution for the satellite pseudo-range residuals within the learning and modeling channel is proposed using the LightGBM algorithm. In addition, a weight decision strategy for the satellite measurements is presented within the weight calculation channel by involving the calibration by the predictions from multiple neighboring vehicles. Results of experiments show that the established pseudo-range residual model using the LightGBM method achieves an enhanced prediction capability over Support Vector Machine (SVM), Random Forest (RF), and LightGBM constructed by basic characteristics. The Root Mean Square Error (RMSE) is reduced by 62.1%, 29.9%, and 60.4%, respectively. The standard deviation of horizontal position error by the proposed solution with the LightGBM model and cooperative weighting is reduced by 48.5% and 47.6% compared with the equal weight strategy and the elevation/SNR (Signal Noise Ratio) integrated weighting strategy. The results illustrate the great value inguaranteeing the advantage of the information interaction mechanism and optimizing the performance of GNSS-based vehicle positioning under the cooperative vehicle infrastructure scheme.