An OFDM receiver sees Doppler and local-oscillator mismatch through the same observable: carrier frequency offset (CFO). For a moving vehicle, that ambiguity makes a one-way CFO estimate insufficient for accurate radial-velocity sensing. A frequency shift may reflect motion, hardware clock drift, or both.
A new public preprint uses the two directions of a 5G NR time-division-duplex link to separate the terms. The method turns ordinary link synchronization measurements into a network-assisted sensing signal.
Uplink and downlink reverse the useful sign
The roadside base station and vehicle estimate CFO from phase-tracking reference signals on downlink and uplink shared channels. Motion-induced Doppler changes sign with link direction, while the relative local-oscillator offset enters the two observations with a different symmetry.
Under the paper’s model, adding the two directional estimates isolates radial Doppler and therefore radial velocity. Taking their difference recovers the relative oscillator offset. The separation uses both directions rather than asking one measurement to distinguish physically confounded effects.
The estimator is closed form and feedback-light
The reported estimator is non-iterative and needs a compact scalar CFO report on the uplink. The authors also discuss scheduling it with a New Radio round-trip-time positioning event, allowing range and radial-velocity measurements to share a link-level procedure.
The paper derives an idealized Cramér–Rao lower bound and reports that a linear phase-slope estimator reaches that bound at high signal-to-noise ratio. It also evaluates sensitivity to angle-of-arrival uncertainty and multipath and discusses mismatch between the two RF chains.
Standards alignment does not replace field validation
The design is aligned with 5G NR PDSCH and PUSCH phase-tracking reference-signal procedures, which makes the signaling path concrete. The remaining question is whether the bidirectional estimates stay sufficiently reciprocal in real roadside deployments with scheduling gaps, acceleration, blockage, independent RF impairments, and rapidly changing multipath.
The public evidence is link-level analysis and numerical evaluation. It does not establish roadside trial accuracy or performance under all implementation and mobility conditions.
Research notes
Bidirectional CFO Separation for Radial Velocity Estimation In 5G NR TDD V2X Links
Authors: Mohamed Elamine Benattia and Huseyin Arslan.
Status: Public arXiv record dated 2 September 2026.
What the public evidence establishes: The work combines uplink and downlink phase-tracking-reference-signal CFO estimates to separate radial Doppler from relative oscillator offset in closed form.
Limits: The available evidence does not establish field-trial accuracy or robustness to every scheduling, mobility, multipath, and RF-chain condition.