V2X: Vehicle-to-Everything Communication (DSRC vs C-V2X vs NR-V2X)

V2X is the direct radio link between vehicles, roadside units (RSUs), vulnerable road users, and the network. It stacks four hard problems in one system: millisecond latency budgets, MAC design without a base-station referee, spectrum policy decided by regulators, and a privacy-preserving PKI at fleet scale. Embedded AI covers the application taxonomy at survey level; this page goes deep on radios, MACs, deadlines, and security, and checks the advertised budgets with a runnable calculator.

The 5.9 GHz fight: why three stacks exist

Every V2X stack traces back to one spectrum decision. In 1999 the FCC reserved 75 MHz at 5.850-5.925 GHz for DSRC (ratified as IEEE 802.11p in 2010); two decades later it concluded the band was mostly idle and re-issued most of it.

• FCC 20-164 First Report and Order, adopted Nov 18, 2020: lower 45 MHz (5.850-5.895) to unlicensed U-NII-4; upper 30 MHz (5.895-5.925) retained for ITS with C-V2X as the required technology; DSRC licensees had one year from the effective date to vacate the lower 45 MHz.
• FCC 24-123 Second Report and Order (adopted Nov 20, 2024; Federal Register Dec 13, 2024): final C-V2X service rules for 5.895-5.925 GHz, after an interim April 2023 waiver decision let deployments use the upper 20 MHz.
• EU: Implementing Decision (EU) 2019/1345 harmonized 5875-5935 MHz for safety-related ITS – technology-neutral, so ITS-G5 (the 802.11p family) and C-V2X coexist politically and physically.
• China: MIIT designated 5905-5925 MHz for LTE-V2X direct communications in Nov 2018 (no station license for in-vehicle/portable units); in Dec 2024 it optimized the configuration to allow 10 MHz or 20 MHz channels.

5.850    5.895      5.905  5.925  GHz  (US band plan after FCC 20-164)
  |........|..........|......|
  | lower 45 MHz      | upper 30 MHz: ITS, C-V2X mandated
  | U-NII-4 unlicensed| deployments use upper 20 MHz (5.905-5.925)
  |                   | after the Apr 2023 waiver decision

EU: 5875-5935 MHz ITS (2019/1345)   China: 5905-5925 MHz LTE-V2X (2018)

Consequence: 802.11p kept the standards, C-V2X kept the spectrum in the US and China, and Europe left the fight unresolved. Everything below is downstream of that sentence.

Message classes and their deadlines

SAE J2735 defines the North American message set; the 2024 revision (J2735_202409, eighth edition) renames the document from “DSRC Message Set Dictionary” to “V2X Communications Message Set Dictionary” – the DSRC-only framing is gone. Europe slots ETSI CAM/DENM into the same architectural roles. China’s T/CSAE 53-2017 industry standard defines the BSM/RSM/SPAT/MAP/RSI equivalents used in its deployments.

Deadlines that matter (each verified against the cited document):

The platooning budget people quote in interviews is 10 ms; the current 5GAA catalog specifies 50 ms for platooning in steady state and keeps 100 ms for collision warnings. Quote the version you read.

Three radio stacks

DSRC: IEEE 802.11p (and 802.11bd)

Wi-Fi halved in frequency domain: 10 MHz channels (the original US plan had seven, channel 178 the control channel), PHY rates 3-27 Mbps, safety frames usually at the QPSK rate-1/2 point (6 Mbps) for robustness. It runs in OCB mode (Outside the Context of a BSS): no association, no MAC-level authentication, every frame broadcast to everyone. Timing doubles vs 20 MHz Wi-Fi: 13 us slots, 32 us SIFS, 58 us DIFS. The MAC is EDCA CSMA/CA; the highest access class (used for safety frames) has CWmin = 3, so uncontended access costs AIFS plus a mean 1.5-slot backoff (~80 us), and contention collapse under load at high density is the well-known weakness. The successor 802.11bd-2022 (published Mar 2023, folded into the IEEE 802.11-2024 base revision) adds dual half-bandwidth mode, LDPC, midambles for high-Doppler tracking, and a preamble that lets 11p and 11bd devices detect each other – but no regulator has handed it new spectrum.

C-V2X LTE sidelink: 3GPP Release 14

Rel-14 (2017) added the PC5 sidelink for direct vehicle-to-vehicle transmission independent of base stations. TS 36.300 (cl. 23.1) defines two resource allocation modes: “Scheduled resource allocation” (eNB schedules, commonly mode 3) and “UE autonomous resource selection” (mode 4), where the UE “performs sensing for (re)selection of sidelink resources” and may keep up to two parallel resource reservations. Mode 4 sensing combines energy detection with control-channel decoding; the chosen resource is held with sensing-based semi-persistent scheduling (SB-SPS), reserved every resource reservation interval (100 ms in the Rel-14 evaluation assumptions) and re-evaluated only occasionally. Rel-14 evaluations targeted 100 ms end-to-end delivery – which the grant cadence itself nearly consumes. Congestion control is CBR-based (channel busy ratio drives power and packet rate).

NR-V2X sidelink: 3GPP Release 16

TS 38.300 (cl. 5.7.2): “Two sidelink resource allocation modes are supported: mode 1 and mode 2. In mode 1, the sidelink resource allocation is provided by the network. In mode 2, UE decides the SL transmission resources in the resource pool(s).” Rel-16 adds unicast and groupcast with HARQ feedback on a new feedback channel (PSFCH), RLC acknowledged mode, and CSI reporting (cl. 16.9) – what platooning and cooperative driving need. Since grants can be configured per slot (sub-millisecond periods), mode 2 removes mode 4’s 50 ms mean grant wait. TR 37.885 supplies the evaluation traffic models: periodic 300/190-byte patterns at 100 ms arrivals with a 100 ms deadline, and a 10 ms-arrival model with a 10 ms deadline.

MAC access patterns for one 10 Hz periodic message

802.11p   each frame: DIFS + CSMA backoff + tx       ~80 us idle; collapses under load
LTE m4    sense... [grant][tx] .. 100 ms .. [grant][tx]     mean wait = RRI/2 = 50 ms
NR m2     CG slot every ~1 ms: [tx][tx][tx]...              mean wait ~0.5 ms

Latency budget anatomy (runnable demo)

End-to-end latency decomposes into radio transmit time (packet/PHY rate), MAC access, modem+stack processing, and propagation (3.34 ns/km – negligible; at 300 km/h a vehicle covers 8.3 cm per ms, so range matters via link budget, not time-of-flight). The demo computes all components for a 300-byte message – the size 5GAA’s forward-collision-warning requirement states, “based on experience from CAM/BSM” – and checks pass/fail against the verified deadlines.

#!/usr/bin/env python3
"""End-to-end latency budget calculator for three V2X radio stacks.

A 300-byte safety message goes through radio hop, MAC access, processing,
and propagation; the sum is checked against published deadlines
(3GPP TR 22.885 V14.0.0; 5GAA SLR Vol I/II v2.0, Nov 2024).

Stack parameters are MODEL INPUTS (typical safety-message operating
points), not standard-mandated constants:
- 802.11p: 6 Mbps (QPSK-1/2 class, 10 MHz); access = DIFS 58 us
  (SIFS 32 us + 2 slots) + mean backoff 1.5 slots x 13 us (CWmin = 3).
- LTE mode 4: sensing-based semi-persistent scheduling; a packet waits
  for the next grant, mean RRI/2 = 50 ms (RRI 100 ms in Rel-14 evals).
- NR mode 2: configured grant per ~1 ms slot; mean wait 0.5 ms; 12 Mbps
  (16QAM-1/2 class).
"""
from __future__ import annotations

C = 299_792_458.0                     # speed of light, m/s
PACKET_BYTES = 300                    # 5GAA FCW information requirement (CAM/BSM class)
TX_PROC_MS, RX_PROC_MS = 0.5, 1.0     # modem + stack pipeline (model input)

STACKS = [
    # (name, PHY rate Mbps, access model, params)
    ("802.11p",    6.0,  "csma", {"difs_us": 58.0, "slot_us": 13.0, "cw_min": 3}),
    ("LTE mode 4", 6.0,  "sps",  {"rri_ms": 100.0}),
    ("NR mode 2",  12.0, "cg",   {"cg_ms": 1.0}),
]

CASES = [
    # (use case, deadline_ms, range_m, source)
    ("Platooning (legacy SLR)",   10.0,  50, "5GAA early SLR: 10 ms / 99.99%"),
    ("Platooning (SLR v2.0)",     50.0,  50, "5GAA Vol II v2.0, Nov 2024"),
    ("Pre-crash sensing",         20.0,  50, "TR 22.885 CPR-015"),
    ("Forward collision warning", 100.0, 150, "TR 22.885 PR.5.1.5-007; 5GAA Vol I 6.1.8"),
    ("V2I signal-phase warning",  100.0, 300, "TR 22.885 CPR-016"),
]


def access_ms(kind, p):
    if kind == "csma":                    # AIFS + mean backoff draw
        return (p["difs_us"] + 0.5 * p["cw_min"] * p["slot_us"]) / 1000.0
    if kind == "sps":                     # wait for next semi-persistent grant
        return p["rri_ms"] / 2.0
    return p["cg_ms"] / 2.0               # wait for next configured grant


def e2e_ms(rate_mbps, kind, p, range_m):
    tx = PACKET_BYTES * 8 / (rate_mbps * 1e6) * 1000.0
    return tx + access_ms(kind, p) + TX_PROC_MS + RX_PROC_MS + range_m / C * 1000.0


def main():
    print(f"Packet: {PACKET_BYTES} B | processing: {TX_PROC_MS + RX_PROC_MS} ms | "
          f"prop: {1e9 / C:.2f} ns/km\n")
    print("Breakdown at 150 m, ms: tx / MAC access / proc / prop")
    for name, rate, kind, p in STACKS:
        tx = PACKET_BYTES * 8 / (rate * 1e6) * 1000.0
        print(f"  {name:<10s} {tx:7.3f} / {access_ms(kind, p):8.3f} / "
              f"{TX_PROC_MS + RX_PROC_MS:5.2f} / {150 / C * 1000.0:.4f}")
    print()
    hdr = (f"{'use case':28s} {'budget':>7s} |"
           + "|".join(f" {name:>17s} " for name, _, _, _ in STACKS))
    print(hdr)
    print("-" * len(hdr))
    for label, budget, rng, src in CASES:
        cells = []
        for _, rate, kind, p in STACKS:
            t = e2e_ms(rate, kind, p, rng)
            cells.append(f" {t:10.2f}ms {'pass' if t <= budget else 'FAIL':<4s} ")
        print(f"{label:28s} {budget:6.0f}ms |" + "|".join(cells))
    print()
    for label, _, _, src in CASES:
        print(f"  {label:28s} <- {src}")


if __name__ == "__main__":
    main()

Output (verbatim):

Packet: 300 B | processing: 1.5 ms | prop: 3.34 ns/km

Breakdown at 150 m, ms: tx / MAC access / proc / prop
  802.11p      0.400 /    0.077 /  1.50 / 0.0005
  LTE mode 4   0.400 /   50.000 /  1.50 / 0.0005
  NR mode 2    0.200 /    0.500 /  1.50 / 0.0005

use case                      budget |           802.11p |        LTE mode 4 |         NR mode 2 
-------------------------------------------------------------------------------------------------
Platooning (legacy SLR)          10ms |       1.98ms pass |      51.90ms FAIL |       2.20ms pass 
Platooning (SLR v2.0)            50ms |       1.98ms pass |      51.90ms FAIL |       2.20ms pass 
Pre-crash sensing                20ms |       1.98ms pass |      51.90ms FAIL |       2.20ms pass 
Forward collision warning       100ms |       1.98ms pass |      51.90ms pass |       2.20ms pass 
V2I signal-phase warning        100ms |       1.98ms pass |      51.90ms pass |       2.20ms pass 

  Platooning (legacy SLR)      <- 5GAA early SLR: 10 ms / 99.99%
  Platooning (SLR v2.0)        <- 5GAA Vol II v2.0, Nov 2024
  Pre-crash sensing            <- TR 22.885 CPR-015
  Forward collision warning    <- TR 22.885 PR.5.1.5-007; 5GAA Vol I 6.1.8
  V2I signal-phase warning     <- TR 22.885 CPR-016

Read the table honestly: the 51.9 ms for LTE mode 4 is not a measurement, it is the structural mean of waiting for the next semi-persistent grant – and jitter around it means the legacy 10 ms platooning budget is out of reach whenever a message misses its grant slot. That is precisely why Rel-16 moved to slot-granular configured grants. 802.11p’s 1.98 ms assumes an uncontended channel; under dense load the CSMA term dominates instead. Processing (1.5 ms) is a model input – real modems add signature verification and filtering there. This budget discipline transfers to real-time systems and to wired in-vehicle networks via TSN.

Security: pseudonym certificates at fleet scale

Every safety message is signed, never encrypted – it is broadcast to strangers – with ECDSA signatures carried in IEEE 1609.2-2022 message and certificate formats. Vehicles hold many short-lived pseudonym certificates that authenticate BSMs without a stable identity; rotation defeats cross-rotation tracking, and enrollment vs authorization duties split across credential types (IEEE 1609.2.1). The regions differ in who runs the trust hierarchy: the US uses the SCMS (Security Credential Management System, the federated CA design documented by the USDOT ITS JPO), Europe runs CCMS with the European Certificate Trust List operated through the Commission’s C-ITS Point of Contact (CPOC). Receivers must also flag attackers and buggy senders, so both architectures add misbehavior reporting – a distributed-detection problem interviews increasingly probe.

Deployment reality by region

• US: C-V2X won by regulation. DSRC vacated the lower 45 MHz after the one-year transition; the April 2023 waiver and FCC 24-123 cemented C-V2X in the upper band. State DOT corridors (transit signal priority, work-zone warnings) migrated DSRC to C-V2X; Audi’s Virginia traffic-light-information deployment was the marquee automaker migration (announced 2020).
• EU: no winner declared. C-Roads pilots run ITS-G5, the Commission kept 5875-5935 MHz technology-neutral, and industry statements keep asking Brussels to commit to a path. With a European supplier, expect both vocabularies (CAM/DENM over ITS-G5, C-V2X trials).
• China: LTE-V2X PC5 at scale via the vehicle-road-cloud program: RSUs on city corridors and highways, the T/CSAE 53 message sets, and the Dec 2024 channel re-configuration (10/20 MHz) as deployments grow.

Summary: the US chose technology by spectrum policy, Europe chose not to choose, and China built the largest deployment on the remaining technology.

Interview drill

• Why did mode 4 struggle with a 10 ms platooning budget? SB-SPS holds a resource every reservation interval (100 ms in Rel-14 evals); mean wait is RRI/2 = 50 ms, so grant cadence alone exceeds the budget.
• What is OCB, and why does 802.11p need it? Outside the Context of a BSS: frames broadcast without association/authentication because every car and RSU is a transient peer; security moves up to 1609.2 signatures.
• What do Rel-16 unicast/groupcast add over broadcast? PSFCH HARQ feedback, RLC acknowledged mode, CSI – reliability for platooning and cooperative maneuvers rather than best-effort awareness.
• Why sign BSMs but not encrypt them? Receiver set is unknown and ad hoc; the goal is authenticity, and privacy comes from pseudonym rotation.
• Where do 100 ms and 10 Hz for forward collision warning come from? TR 22.885 PR.5.1.5-007/-008 and the 5GAA SLR (100 ms, 99.99%, 300 B, 150 m).
• Where does V2X meet edge computing? V2N2V paths and RSU sensor fusion (RSM/RSI) terminate on multi-access edge nodes – see Edge Computing and MEC latency tiers in 5G.

References

1. FCC, First Report and Order, ET Docket 19-138, FCC 20-164 (Nov 18, 2020) – https://docs.fcc.gov/public/attachments/FCC-20-164A1.pdf (probed: OK)
2. FCC, Second Report and Order, FCC 24-123, Federal Register (Dec 13, 2024) – https://www.federalregister.gov/documents/2024/12/13/2024-28980/use-of-the-5850-5925-ghz-band (probed: OK)
3. 3GPP TR 22.885 V14.0.0, Study on LTE support for V2X services – https://www.3gpp.org/ftp/Specs/archive/22_series/22.885/22885-e00.zip (probed: OK)
4. 3GPP TS 36.300 V14.0.0, E-UTRA and E-UTRAN Overall Description, cl. 23.1 – https://www.3gpp.org/ftp/Specs/archive/36_series/36.300/36300-e00.zip (probed: OK)
5. 3GPP TS 38.300 V16.6.0, NR and NG-RAN Overall Description, cl. 5.7, 16.9 – https://www.3gpp.org/ftp/Specs/archive/38_series/38.300/38300-g60.zip (probed: OK)
6. 3GPP TR 37.885 V15.0.0, NR V2X evaluation methodology – https://www.3gpp.org/ftp/Specs/archive/37_series/37.885/37885-f00.zip (probed: OK)
7. 5GAA, C-V2X Use Cases and Service Level Requirements, Vol I / Vol II v2.0 (Nov 2024) – https://5gaa.org/publications/c-v2x-use-cases-and-service-level-requirements/ (probed: OK)
8. SAE J2735_202409, V2X Communications Message Set Dictionary – https://www.sae.org/standards/j2735-v2x-communications-message-set-dictionary (sae.org blocks scripted probes; verified via SAE Mobility and ANSI listings)
9. IEEE 802.11bd-2022, Amendment 5: Enhancements for Next Generation V2X – https://standards.ieee.org/ieee/802.11bd/7451/ (standards.ieee.org returns 403 to scripted probes; verified via IEEE Xplore, published Mar 2023)
10. IEEE 1609.2-2022, Security Services for Applications and Management Messages – https://standards.ieee.org/ieee/1609.2/10258/ (same bot-wall note)
11. Commission Implementing Decision (EU) 2019/1345 (5875-5935 MHz ITS) – https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32019D1345 (probed: OK)
12. USDOT ITS JPO, Security Credential Management System – https://www.its.dot.gov/resources/scms.htm ; EC C-ITS Point of Contact – https://cpoc.jrc.ec.europa.eu/ (both probed: OK)