Stochastic Geometry-Based Modelling and Analysis of Vehicular Networks: Supporting V2V and Visual Collaborative Environment Sensing

Vision-based environment perception is a crucial component in autonomous driving technology, predominantly utilizing cameras and other visual sensors to gather information from the surrounding environment. However, the perception capability of a single vehicle is limited. To address this issue, in this paper, we propose a V2V-enabled collaborative visual environment perception framework to enhance perception accuracy and reliability of a single vehicle through information sharing and collaboration between vehicles. Particularly, we utilize the transdimensional Poisson point process (TPPP) to model the actual distribution of vehicles, where TPPP is a combination of 1D and 2D PPPs. We first derive the coverage probability of a typical vehicle located at the origin. Then, assuming that each vehicle has a reliable perception distance <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$d$</tex>, we quantify the fraction of the mean area covered from successfully collaborating vehicles within a circular region of radius <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathcal{D}$</tex> centered at the origin. It is worth noting that the metric of mean area fraction fundamentally reflects the effectiveness and performance of V2V-enabled and vision-based collaborative environmental perception. Finally, the analytical findings of this paper are validated through simulation experiments.