C-V2X Cooperative Positioning and Multi-Source Fusion Control for Autonomous Driving
Abstract
Reliable centimeter-level position measurement is essential for advanced autonomous driving. In complex urban environments, if the Global Navigation Satellite System (GNSS) fails to operate, C-V2X cooperative positioning provides a reliable alternative solution, but it faces limitations such as channel confusion and physical layer distance measurement. This thesis will analyze and study the most advanced V2X cooperative positioning technologies based on the "communication- computation-control" framework. We believe that in order to overcome the bottlenecks of signals and errors, a deep cross-layer closed-loop is required. Specifically, the downstream layer stacks deep learning, reinforcement learning and uses multiple access technology for resource allocation. In the middle, in the neural network for information transmission at the layer level, non-line-of-sight (NLOS) errors are applicable and the field of view (FOV) is widened. The upper layer stacks on roadside units (RSUs) by using dynamic errors to guide the execution of adaptive fusion filtering, enabling micro-level lane change control to be achieved. This comprehensive paper emphasizes the urgency of the deep integration of communication networks and multi-sensor fusion, and suggests the direction towards an integrated perception and communication (ISAC) architecture in 6G.