AI + V2X cooperative infrastructure

Traffic Perception, Prediction, Optimization, Coordination, and Connectivityfor safer, smoother roadways

Kemetra turns intersections into cooperative systems: perceiving every road user at the edge, predicting conflicts before they happen, clearing queues before they build, and speaking directly to the vehicles around them.

Request a pilotSee how it works

LIVE AT THE EDGE<100 MS SENSE → BROADCASTDIRECT 5.9 GHz C-V2XAUSTIN, TX

The streets we’re building toward

Zero crashes.

Zero congestion.

Zero emissions.

Every cooperative intersection moves a city closer: preventing the conflict, clearing the queue, and cutting the emissions that stop-and-go traffic leaves behind.

How it works

Sense. Decide. Deliver.

01 SENSE

Every road user, perceived at the corner.

vEdge perceives every road user on-device at the corner, day, night, rain. No cloud round-trip. No raw video ever leaves the intersection.

02 DECIDE

Conflicts predicted seconds ahead.

It predicts conflicts seconds before they happen and reallocates green time continuously, at the edge, so it keeps working even when the network doesn't.

03 DELIVER

Decisions land on the street.

In milliseconds: protective signal actions, priority for emergency vehicles and transit, and standards-based C-V2X messages to every equipped vehicle.

vEDGE REQUESTS. THE CABINET DECIDES. THE MMU CONFLICT MONITOR REMAINS THE CERTIFIED SAFETY AUTHORITY, ALWAYS.

vEdge · Hardware

Perception, C-V2X, and control. One node in the cabinet.

An NVIDIA Jetson-based roadside node: 200 TOPS of on-device perception fused with a direct 5.9 GHz C-V2X radio. It detects and classifies every road user in real time, speaks the connected-vehicle standards, and drops into the signal cabinet a city already owns.

  • JETSON AGX ORIN · 200 TOPS
  • DUAL 5.9 GHz C-V2X RADIOS
  • CARRIER I/O · NTCIP · RTSP
  • NEMA TS2 CHASSIS

PLATFORM · NEMA TS2 CERTIFIED (FCC · CE)
V2X MODULE · OMNIAIR-CERTIFIED LTE-V2X PC5

The Kemetra K-RSU roadside compute node: a fanless industrial unit with a thick finned aluminum heatsink over a matte charcoal chassis, and a single flush front panel carrying ethernet, USB, terminal-block and C-V2X antenna connectors.

vMobility · Software

From the corner to the control room.

Every vEdge streams up to vMobility: one live operational picture of the whole corridor. It fuses spatial intelligence with real-time edge inference from every corner, closing the gap between what the map knows and what the street is doing right now. Signal performance measures your engineers already use, safety analytics that surface near-misses before they become crashes, and reporting formatted for the funding programs that pay for safer streets.

  • WORLD CONTEXTthe built environment around the corridor
  • STATIC TWINlane geometry, signals, and approaches
  • LIVE TWINreal-time edge inference from every corner
  • ANALYTICAL TWINsafety, performance, and reporting
vMobility operations view on a studio monitor: a live 3D map of downtown Austin with device status for 120 cameras and 30 vEdge nodes, mobility health scores by district, congestion trend, incident alerts, and air quality.

Solutions

Three problems. One node.

INTERSECTION SAFETY

Predict turn conflicts. Prevent the crash.

Detection is table stakes; acting in time is the product. vEdge scores every interaction between road users, predicts conflicts seconds ahead, and acts: extending pedestrian clearance, holding a conflicting phase, broadcasting a warning to equipped vehicles. The same node runs automated incident detection at the corner, flagging crashes, heavy congestion, flooded roadways, and active work zones the moment they appear. Every event is captured with evidence for engineers and funding applications.

Conflict predictionAutomated incident detectionActive protectionVulnerable road user safetyNight & adverse-weather fusionEvidence capture
A signalized intersection where a blue through car brakes as a silver sedan turns left across its path; the vEdge has predicted the conflict and thrown a red warning alert at the crossing point, its amber turn path crossing the car's teal through path.
  • TTC 2.4 s · RISK 0.87left turn cuts across the through path
  • VEHICLE WARNED · COLLISION AVOIDEDthe node acts seconds ahead
  • C-V2X 5.9 GHzwarning broadcast · evidence logged

SIGNAL OPTIMIZATION

Green time that follows real demand, not a fixed timing plan.

vEdge measures real queues, pedestrians, and demand on-device; vMobility reallocates green time continuously, cutting delay and clearing the path for emergency vehicles the moment they roll. It coordinates neighboring signals into a green wave and gives buses priority so a late route makes the light, not the next cycle. Crossing time is dynamic too: when the node sees a slower pedestrian, a wheelchair, or a crowd still in the crosswalk, it extends the walk interval so everyone clears before the light turns.

Adaptive controlCorridor coordinationDynamic crossing timeEmergency preemptionTransit priorityQueue & demand sensing
A busy four-way city intersection where the signal facing the arterial has been given a green and a transit bus in the right-hand lane rolls through with the traffic; the vEdge holds the green for the corridor so the bus and the cars are not stopped.
  • TRANSIT GREENWAVETSP held across the corridor
  • BUS ROLLS THROUGHgreen wave, multiple intersections
  • DIRECT C-V2X · 5.9 GHzon-device, no cloud relay

V2X & AV INTEGRATION

The intersection becomes another sensor for every connected vehicle.

Kemetra broadcasts signal phase and timing (SPaT/MAP) and cooperative perception (SDSM) straight to autonomous and connected vehicles over direct 5.9 GHz C-V2X, giving each one non-line-of-sight awareness of the road users, hazards, and signals its own sensors can’t see yet. When vEdge detects flooding, a crash, or a road-closing event ahead, it feeds that to approaching AVs so they can reroute before they reach it. Sub-100-millisecond, standards-based, engineered to the national Connected Intersection guide, not a cloud relay.

Signal phase & timingCooperative perceptionNon-line-of-sight (NLOS)Hazard reroutingBlind-corner warningsEngineered to CTI 4501

Built in Austin, where autonomous fleets already share the road every day.

An autonomous vehicle in the curbside lane brakes at a red signal before a right turn. A pedestrian approaching the corner is hidden behind the building, out of the vehicle's sight, and is shown by the roadside sensor as a glowing teal detection silhouette through the building corner, illustrating non-line-of-sight perception.
  • NLOS DETECTIONsees the pedestrian the vehicle can't
  • SDSM SENT · AV WARNEDcooperative perception over C-V2X
  • 84 ms LATENCYdirect 5.9 GHz, no cloud relay

Deployment

Built for the cabinet you already own.

Start with detection. Unlock safety and V2X by software.

EXISTING CONTROLLERS

NTCIP 1202 · no rip-and-replace

EXISTING CAMERAS

reuses qualified city CCTV (RTSP/ONVIF)

ONE-DAY INSTALL

one truck-roll, no bucket truck

PRIVACY BY DESIGN

all inference on-device · no raw video, no PII, no facial recognition

A signalized intersection deployment: an existing traffic camera and a C-V2X antenna kit on the mast-arm pole, with the compact K-RSU node mounted inside the open roadside signal cabinet.

Applications

Intersections

Safety, signal timing, and V2X from one node at the corner.

Curbside & AV pickup/drop-off

The same node manages the curb: who stops, where, and for how long, all in software.

Parking & Occupancy

Occupancy and enforcement from infrastructure already in place.

36,640

estimated U.S. traffic deaths in 2025, concentrated at intersections and among vulnerable road users.

SOURCE · NHTSA, 2026

~50%

of injury crashes occur at or near intersections.

SOURCE · FHWA

<100 ms

target end-to-end latency, perception → broadcast, on-device.

KEMETRA DESIGN TARGET

Kemetra exists to bend the first two numbers down: every conflict captured with evidence, in the reporting format Vision Zero and federal safety programs already use.

Built by

A team that has shipped self-driving systems, AI compute, and hardware at global scale.

Tesla · General Motors · NVIDIA · HP · Dell

The intelligent intersection starts today.

We partner with cities, DOTs, and research programs on pilot corridors. Tell us about your intersection and we’ll scope a deployment: hardware, software, and the evidence plan.