Confidence-Calibrated Positioning for Uncertainty-Aware V2X Systems
Year: 2026
Type: conference-paper
Abstract: Reliable localization is critical for cooperative vehicular systems, yet most methods optimize accuracy without quantifying positional trust. This work proposes a hybrid 5G–Global Navigation Satellite Systems (GNSS) framework that attaches a calibrated confidence score to each position estimate, enabling uncertainty-aware operation in Vehicle-to-Everything (V2X) environments. Fusing 5G radio observables with GNSS data through adaptive gain filtering and probabilistic mapmatching reduces median error from 27 m to below 10 m in a live 5G testbed, with a confidence score empirically calibrated against absolute positioning error$(r=-0.59)$, providing a monotonic estimate of positional reliability. This interpretable confidence layer enables vehicles and infrastructure to adapt their behavior based on positional trust, improving both safety decisions and communication efficiency in cooperative scenarios. By exposing confidence as a signal, the framework bridges localization, communication, and cooperative decision-making in future 6G V2X systems, validated against synchronized GNSS ground truth over approximately 45,000 samples.
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7202363316