V2X Primer — Connecting Vehicles, Roads, Networks, and People Safely

At a blind intersection, a cyclist can be hidden behind a building from the ego vehicle's camera and radar. A roadside sensor or another vehicle may see the cyclist first and transmit position and motion. That message can buy seconds of prediction time. V2X, or vehicle-to-everything, does not magically extend a sensor's line of sight. It shares someone else's observation and intent, with an age, uncertainty, and expiration time.

A design in which communication success means safe and packet loss means unsafe is incomplete. Radio links are blocked, channels become congested, clocks drift, maps disagree, and certificates expire. V2X should therefore contribute evidence to perception and planning while the vehicle remains able to degrade safely when that evidence disappears.

Example of a vehicle combining onboard sensing with connected functions

Image: Subaru WRX S4 2.0GT-S EyeSight (Tokumeigakarinoaoshima, CC BY-SA 4.0), Wikimedia Commons. This is an exterior vehicle view, not evidence of a particular V2X radio, protocol, or antenna layout.

The short conclusion

• V2V connects vehicles, V2I connects road infrastructure, V2N reaches cellular networks and cloud services, and V2P includes pedestrians and cyclists. Their deadlines and trust boundaries differ.
• DSRC/IEEE 802.11p families provide direct ad-hoc exchange. C-V2X combines direct PC5 sidelink with Uu connectivity through a cellular network.
• Mean latency is inadequate for safety. Measure deadline delivery, tail latency, packet delivery ratio by distance and density, burst-loss length, and Age of Information.
• A valid signature proves possession of an authorized key and message integrity. It does not prove that a GNSS position is physically true or that a legitimate sender is honest.
• Frequent position broadcasts create a tracking risk. Pseudonym certificates help, but timing and trajectory can still link successive identities.
• Communication loss must not create an unsafe command. Increase uncertainty and return smoothly to an onboard-sensor-only behavior.

The four communication relationships

Figure 1 — “X” is not one radio standard. Direct and network paths serve different time horizons and failure assumptions.

V2V: exchanging motion and intent

Vehicles can broadcast position, speed, acceleration, heading, brake state, dimensions, and path intent for emergency-braking and intersection warnings. A received position still contains GNSS, clock, and frame-transform error. Treat it as a timed probabilistic observation, not as a perfectly located obstacle.

V2I: connecting signal controllers and road sensors

Roadside units can publish signal phase and timing (SPaT), intersection geometry (MAP), work zones, speed restrictions, surface warnings, or emergency-vehicle priority. The source signal controller, RSU, map version, and clock form one data chain. A fresh radio packet containing stale controller state is still stale.

V2N: reaching wide-area services

Cellular access supports congestion ahead, map updates, software delivery, fleet operation, and cloud-assisted computation. Its path crosses a radio access network, core, and application service. Collision-imminent control should not rely solely on a remote server whose route and availability the vehicle cannot bound.

V2P: including vulnerable road users

A phone or dedicated device may announce a pedestrian or cyclist. Power-saving update rates, poor placement inside a bag, and urban GNSS multipath can put the reported point on the wrong side of a road. Cross-check it with roadside sensing and the vehicle's monocular camera, while controlling nuisance-alert rate.

DSRC/ITS-G5 versus C-V2X

DSRC systems based on IEEE 802.11p use Wi-Fi-related contention methods adapted for rapidly changing vehicular links. They can exchange short messages without first associating with a conventional access point. Europe commonly refers to the family as ITS-G5, while WAVE names a related US standards stack. Its decentralized operation is attractive, but dense traffic raises collisions, hidden-node effects, and channel-load problems.

C-V2X does not mean every packet traverses a cellular base station:

• PC5 sidelink supports direct device-to-device exchange, evolving from LTE-V2X toward NR-V2X.
• Uu connects through cellular radio access, a core network, and services for wide-area information and management.

The label does not determine field performance. Antenna placement, vehicle-body blockage, channel bandwidth, transmit power, modulation, resource selection, retransmission, security overhead, and traffic density all matter. Compare application deadline success under the same distance, obstruction, and load rather than comparing acronyms in isolation.

Latency, reliability, and information age

For message generation at t_s and application availability at t_a, end-to-end latency is

A mean E[L] hides rare 500-ms delays. Report P(L\le L_{max}) and high percentiles, and separate clock error from actual transport delay. With N_t transmitted packets and N_r packets received before their deadline,

The same PDR can arise from isolated losses or a ten-packet outage. Report burst-loss length and PDR versus range, speed, obstruction, density, and offered load.

Age of Information measures how old the newest available source sample is. If u(t) is its generation time,

Low air-interface latency cannot repair a measurement that waited in perception or a server queue. Instrument sensor capture, state estimation, signing, queueing, radio, verification, and fusion on a common time base.

For closing distance d and relative velocity v_r<0, a simple time to collision is

An information age \Delta adds approximately |v_r|\Delta of motion uncertainty before acceleration and localization errors. This connects network metrics to the safety margin that planning actually consumes.

Cooperative perception without double counting

Suppose a remote source reports z_A with covariance P_A and ego sensing reports z_E with P_E. Under an independence assumption,

If both estimates originate from the same roadside camera or map, they are correlated and this formula becomes overconfident. Track provenance, source clock, coordinate frame, and known common inputs. Sharing can occur at several levels:

1. Ego state and intent: compact but cannot describe a third party observed by the sender.
2. Object tracks: practical positions, classes, velocities, and covariances, but track IDs must be associated.
3. Features or raw sensor data: potentially rich, but expensive in bandwidth, privacy, model compatibility, and calibration.

For the local state estimate, see Dead Reckoning Primer. For onboard evidence, see LiDAR and automotive radar.

Security: authentic does not necessarily mean true

Vehicular PKI signs messages. Verification can establish that an authorized private key signed the bytes and that the bytes were not changed. It cannot establish physical truth. A legitimate but compromised vehicle can report a false position; a GNSS-spoofed unit can honestly sign its wrong estimate.

Threats include replaying an old valid hazard, flooding the channel or verification CPU, creating many apparent identities, stealing keys through a service interface, and compromising an RSU or update system. Defenses combine signature and certificate checks with timestamp windows, sequence rules, physically possible acceleration, road-map consistency, multi-source agreement, and onboard-sensor comparison. Reporting and revocation need evidence thresholds because falsely excluding a legitimate vehicle is also harmful.

A complete key lifecycle covers a hardware security module, secure boot, signed software, provisioning, pseudonym rotation, revocation, service replacement, ownership transfer, and decommissioning. Cryptographic algorithm choice is only one line in that lifecycle.

Location privacy

A fixed identifier attached to periodic position messages reveals home, work, and routine even without a name. Pseudonym certificates reduce direct linking, but a smooth trajectory, timing, vehicle dimensions, and radio fingerprint can link the old and new identities.

Minimize the data first: do not transmit history that a safety application does not need, aggregate at roadside infrastructure where possible, bound retention, audit access, and specify purpose. Coordinated pseudonym changes or mix zones may reduce trajectory linking. Removing a database ID before publishing a research dataset is not sufficient anonymization.

Safety degradation and evaluation

Every received item needs a quality flag and expiration. When messages stop, do not delete the object as if it vanished; propagate uncertainty:

Evaluate tunnels, urban canyons, heavy-vehicle blockage, rain, coverage loss, dense-channel load, certificate-verification overload, clock errors, map-version disagreement, duplicated objects, replay, malformed and valid-format floods, and the transition between connected and unconnected behavior.

Radio PDR is not the final safety result. Measure collision avoidance, minimum TTC, missed and nuisance warnings, induced harsh maneuvers, degradation success, and unintended risk to following or adjacent traffic. The 2024 US DOT deployment plan similarly treats interoperability, security, privacy, and safety as deployment conditions, not optional additions.

Reproducible implementation sequence

1. Synchronize GNSS/IMU and vehicle-bus measurements to a traceable clock.
2. Attach source timestamp, frame, accuracy, state, pseudonym certificate, and signature.
3. Validate syntax, signature, time window, geographic plausibility, and kinematics at reception.
4. Transform into a map frame and resolve duplicate tracks and correlated sources.
5. Fuse only when quality meets the application contract; retain an onboard-only fallback.
6. Log loss, stale data, security rejection, compute overload, degradation, and recovery.

Progress from software simulation through HIL, closed-course radio scenarios, and controlled field deployment. A link can pass every RF test while an incorrect axis convention places an object across the centerline.

Summary

V2X is not simply “put 5G in a car.” It is a distributed claim about who observed what, when, in which frame, with what uncertainty, and how a receiver may safely use it. Select direct and network paths by deadline, measure tail behavior and information age, verify both cryptographic and physical consistency, and preserve safe behavior without connectivity. Done well, connectivity reveals occluded hazards and shares road intent. Done poorly, it distributes plausible stale coordinates at radio speed.

References

• US DOT — Saving Lives with Connectivity: National V2X Deployment Plan
• 3GPP — Vehicle-to-everything technologies
• ETSI — Automotive Intelligent Transport Systems
• Car 2 Car Communication Consortium
• NHTSA — Vehicle Cybersecurity
• UNECE — UN Regulation No. 155: Cyber security and cyber security management system