Artificial Intelligence-Based Methods for Data Detection in High-Mobility Vehicular Networks: A Systematic Literature Review

Data detection in high-mobility vehicular networks operates under doubly dispersive channels, where Doppler-induced frequency variations limit conventional detectors and motivate the use of artificial intelligence. This systematic review characterizes the current state of AI-based data detection in these systems and identifies research gaps. The review follows the PRISMA 2020 statement, complemented by the guidelines of Kitchenham and Charters. IEEE Xplore, Scopus, and SpringerLink were searched from 2017 to June 2026 for English journal and conference papers on AI-based detection in vehicular receivers. Of 1850 identified records, 166 duplicates were removed, and screening by title, abstract, full text, and methodological quality resulted in 59 primary studies, synthesized narratively. Orthogonal Time Frequency Space (OTFS) is the dominant modulation scheme, present in 67.8% of the studies and in 61.3% of the 31 studies evaluated on emulated or standardized vehicular channels. Deep and convolutional neural networks are the most frequent architecture families, with a growing trend toward Transformer-based methods since 2024. Evidence is limited by heterogeneous simulation setups and insufficient experimental validation. The main gaps are the absence of real-world vehicular tests, minimal treatment of hardware impairments, and few direct comparisons between Orthogonal Frequency Division Multiplexing (OFDM) and OTFS under standardized conditions.