Robust Bayesian Cooperative Positioning for Intelligent Vehicles Using GNSS and V2V Range Measurements

GNSS-based cooperative positioning offers advantages such as high accuracy, robustness, and availability, making it highly effective in enhancing the positioning performance of intelligent vehicles in urban environments. Due to the complex signal propagation conditions in urban settings, GNSS and inter-vehicle measurements often include uncertainties caused by non-ideal factors. These uncertainties introduce anomalous measurement biases and noise with unknown characteristics, degrading positioning accuracy. To address this issue, this paper proposes a robust distributed Bayesian cooperative positioning algorithm. We first introduce latent variables to characterize unknown uncertainties in GNSS and V2V measurements. These latent variables are modeled using Gaussian-Gamma conjugate distributions, with the shape of the distribution determined by hyperparameters. Based on the Variational Bayesian (VB) theory, we then decompose the robust cooperative positioning problem into an alternating estimation of vehicle states and measurement uncertainties. We derive message-passing-based closed-form solutions for updating the variational posteriors of vehicle states and latent variables in a distributed manner, allowing all parameters to be estimated algebraically. Additionally, the computational complexity and communication overhead are also analyzed. Performance evaluation results using datasets from real urban environments show that the proposed algorithm achieves higher positioning accuracy compared to existing methods and is more robust to anomalous measurements. Furthermore, the proposed algorithm is insensitive to nominal parameter settings, featuring low computational complexity and communication overhead.