An Integrated Framework for Vision-Based Relative Localization and Bandwidth Allocation

Abstract

The provision of an accurate and efficient localization system is of great significance to Internet of Vehicles (IoV). In view of its promising applications including formation control and autonomous driving, the vision-based localization mechanism has attracted intensive attention for its outstanding performance in terms of accuracy and flexibility. However, the high mobility and fast switching features of IoV put a strain on communication resources, which can impair the real-time transmission of high-quality images for cooperative localization. In this paper, we propose ajoint position sensing and bandwidthconstrained communication framework for vision localization. First, we derive the relative squared position error bound (SPEB) by implementing subspace projection on the Fisher information matrix (FIM) of absolute positions given quantized observations. Then we exploit the relative geometry of vehicles and images to develop two localization algorithms, which estimate the relative positions of vehicles by incorporating the idea of Euclidean distance matrix. The effectiveness of the proposed localization algorithms is verified via numerical simulations, which offers guidelines for joint sensing and communication in future IoV scenarios.