XShield — AI Accident Reconstruction & Exoneration

The problem

In a crash, the truck is presumed guilty — and the evidence disappears fast.

Nuclear verdicts against carriers keep climbing. Memories shift, witnesses leave, dashcam clips get overwritten, and the FNOL drags on for weeks. By the time anyone reconstructs what really happened, the narrative is already set against your driver.

PRESUMED FAULT

The big truck takes the blame

Juries assume the larger vehicle is at fault. Without hard physics, "he said / she said" favors the other car.

EVIDENCE DECAY

Proof vanishes in hours

Skid marks wash away, video loops over, and the scene is cleared before an adjuster ever sees it.

SLOW FNOL

Claims drag, reserves balloon

Manual first-notice-of-loss takes days. Every day of uncertainty inflates the reserve and the settlement.

NO RECONSTRUCTION

You can't show what happened

A police diagram isn't a reconstruction. Defending the driver means proving the geometry, speed and forces.

How it works

One collision triggers an autonomous chain of evidence.

The moment the telematics device flags an impact, XShield's incident commander runs the entire pipeline on its own — and stops at the one place that matters: a human approves anything that leaves the building.

Detect

The collision rule fires on the Geotab device. The clock — and a litigation hold — starts inside 60 seconds.

Capture

The 100 Hz accelerometer burst, GPS trail and camera footage are pulled and hashed, immutable.

Reconstruct

Sensor fusion + map-matching rebuild the true trajectory, impact point, Δv and direction of force.

Reenact

A photoreal 3D reenactment plays back the crash on the real intersection, at the real time of day.

Compile

Every piece of evidence — favorable and not — assembles into one cited, reproducible dossier.

Exonerate

On your approval, the defense package goes to your insurer and counsel. Subrogation begins.

Case studies

What XShield proves, crash by crash.

Every reconstruction separates what was measured from your truck's own sensors from what is inferred about the other vehicle from physics and video — so the evidence holds up under cross-examination.

MEASURED · your truckINFERRED · other vehicle

Rear-endCASE 01 · cut-in + brake-check

Semi rear-ends a car

The car changed lanes into the gap ahead of the truck and braked hard. Presumptively the truck's fault — until the data shows otherwise.

XShield proves

The car entered with under two seconds of headway and decelerated abruptly, leaving no safe stopping distance. The driver braked on reaction.

following gap 1.4 s at cut-in · ego brake reaction 0.6 s
lead decel 0.74 G · Δv 4.1 m/s · PDOF 12 o'clock

▶ ROAD-CAM corroborates · ▶ DRIVER-CAM attentive

Failure to yieldCASE 02 · left across path

Car turns left into the truck's path

An oncoming car turned left across the intersection directly into the truck travelling straight through on a green.

XShield proves

The truck held the right of way, was within the posted limit, and had no time or distance to avoid the turning vehicle.

ego speed 38 mph in 40 zone · throttle off, no brake possible
time-to-collision 0.9 s · Δv 6.3 m/s · PDOF 1 o'clock

▶ ROAD-CAM shows the turn · ▶ DRIVER-CAM attentive

Weather / surfaceCASE 03 · loss of traction

Icy road — trailer jackknifes

On a highway grade the trailer broke traction and swung out, the rig jackknifing across the lane. No second vehicle — but the carrier is still on the hook.

XShield proves

Speed was appropriate for conditions and braking input was correct; loss of traction was a road-surface event, not driver error. Weather and surface are causation, on the record.

speed 41 mph · drive-wheel slip onset t−1.1 s
steering correction logged · yaw rate 34°/s · no harsh input

▶ ROAD-CAM shows ice + spray · ▶ DRIVER-CAM attentive

Unsafe lane changeCASE 04 · merge into occupied lane

Car merges into the truck's side

A car changed lanes into the space the truck already occupied, striking the tractor's side, then claimed the truck drifted.

XShield proves

The truck held a straight, centered line with zero lateral input; the impact came from the adjacent lane. The lane-keeping data refutes the drift claim outright.

ego lateral deviation 0.0 m · steering centered
impact bearing 9 o'clock · Δv 2.4 m/s · video attached

▶ ROAD-CAM shows the merge · ▶ DRIVER-CAM attentive

Rear-endedCASE 05 · struck from behind

Car rear-ends the truck

The truck was slowing in heavy traffic when a car struck it from behind, then alleged the truck "stopped short."

XShield proves

Brake lights were on and the truck decelerated smoothly within normal limits; the following car closed at speed. A textbook following-too-close impact.

ego decel 0.3 G smooth · brake lights ON · stopped legally
other closing 24 mph · Δv 5.1 m/s · PDOF 6 o'clock

▶ REAR ROAD-CAM corroborates · ▶ DRIVER-CAM attentive

Red-light runnerCASE 06 · ran the light

Car runs a red light

The truck entered the intersection on a green when a car ran the cross-street red and broadsided it.

XShield proves

Entry speed and timing show the truck crossed legally on green with no warning; the other vehicle never decelerated for the red.

ego speed 31 mph on green · no brake possible
other: no deceleration · Δv 6.8 m/s · PDOF 3 o'clock

▶ ROAD-CAM shows the signal · ▶ DRIVER-CAM attentive

Data + vision

The data says what happened. The video proves it.

Telemetry and cameras run on one clock. Every g-force spike lines up with the exact frame that caused it — so the forces your truck felt and the scene your camera saw are the same instant, from two independent sources that can't contradict each other.

EVIDENCE SYNC · road-cam + driver-cam + 100 Hz telemetry · one timeline

📹

Road-facing camera

Shows the other vehicle in the seconds before impact — running the light, cutting in, crossing the line. The behavior, on video.

👁️

Driver-facing (DMS)

Proves your driver was attentive: eyes on the road, hands on the wheel, no phone. Often the single strongest piece of exoneration evidence.

⏱️

One timeline

Every frame is stamped to the same 100 Hz clock as the sensors. The g-force spike and the footage are provably the same moment.

🤖

AI vision events

GO Focus and LightMetrics auto-flag distraction, following distance and lane departure — corroboration that's measured, not argued.

🔐

Tamper-evident

Every clip is hashed at capture and locked to the telemetry. Any edit breaks the hash — so the footage holds up as authenticated evidence.

📡

Secured in seconds

Video and sensor data are pulled and backed up within the first minute — before the loop overwrites it or the scene is ever cleared.

3D reenactment

See the crash replayed on the real street, at the real time of day.

XShield renders the reconstruction on Google's photorealistic 3D map of the exact location — with the true sun position, a scrubbable timeline, and synced telemetry. A jury doesn't read a report. They watch it happen.

RECONSTRUCTION · synthetic incident · MEASURED ego / INFERRED other

FNOL automation · TruckerCloud

From impact to a filed claim — in under 5 minutes.

The instant a crash is detected, XShield assembles the First Notice of Loss automatically — vehicle, driver, telemetry, weather and video on one clock — and routes it straight to TruckerCloud and your carrier. The claim starts at the moment of impact, not whenever the driver gets around to reporting it.

Triggered by the crash, not the driver. Impact detection starts the FNOL automatically — no phone call, no late or missed reports, no "I forgot to mention it."
Automatic to TruckerCloud & carrier. The FNOL and crash alert flow straight into TruckerCloud's insurer network and notify your carrier in real time — within minutes of impact.
ACORD-format packet with the full evidence bundle attached — road-cam, driver-cam and 100 Hz telemetry.
Any ELD, any camera. TruckerCloud reaches carriers already on its network — no building one insurer API at a time.

See it on your fleet →

INCIDENT COMMANDER · crash → FNOL → TruckerCloud

<60simpact → carrier notified

0driver reports required

100+ELD & camera sources via TruckerCloud

Capabilities

An evidence engine, not an advocacy engine.

XShield collects everything — including the inconvenient — because a defense that occasionally says "here is the honest problem, early" is the one juries and adjusters believe.

⚡

100 Hz sensor fusion

Accelerometer, gyro and GPS fused and map-matched into the true sub-second trajectory — not a guess, a reconstruction.

🌐

Photoreal 3D reenactment

The crash replayed on Google's 3D map of the real intersection, at the real time of day, with the real sun and shadows.

⚖️

Measured vs inferred

Your truck's motion is measured; the other vehicle is clearly labeled as inferred. Honesty is what survives cross-examination.

📁

Court-ready dossier

Every exhibit cited, hashed and reproducible from the raw sensor data. An expert can trace any number back to its source.

🛡️

Chain of custody

Raw capture is immutable and hashed within the first minute. A litigation hold starts before anyone touches the scene.

🤖

Autonomous, human-gated

The pipeline runs itself, but nothing reaches an insurer, adjuster or court without an explicit human approval.

📊

Driver scorecard

Always-on, archetype-aware safety scoring that predicts risk before a crash — and never penalizes a driver XShield exonerated.

🔁

Subrogation & FNOL

Automated first-notice-of-loss and a subrogation package that pursues recovery from the genuinely at-fault party.

Defensibility

If it isn't reproducible, it isn't evidence.

Polish doesn't win cases — traceability does. Every XShield reconstruction is a versioned derivation from immutable sensor data. Re-run it and you get the identical result, every time.

Reproducible. Re-render from the same sensor data → the same exhibit, byte for byte.
Explainable. Deterministic physics, not a black box. Every figure traces to a reading.
Complete. Favorable and unfavorable evidence both collected — suppression is spoliation.
Honest. Confidence scored throughout; inferred is never dressed up as measured.

DOSSIER · INTEGRITY MANIFEST

incident_id   INCIDENT-0001
method      recon-2026.06-ekf
inputs_sha256 a17f…9c4e ✓
evidence    13 items · 11 favorable · 2 noted
exhibit     3D · reproducible ✓
fault       suggested, human-reviewed
release     pending approval ⏸

Integrations

It runs on the hardware your fleet already has.

TELEMATICS

Geotab GO9 / GO10

AI-powered collision detection and the 100 Hz accelerometer burst that anchors every reconstruction.

VISION

GO Focus & LightMetrics

Road- and driver-facing AI cameras add video corroboration and attentiveness evidence to the dossier.

NETWORK

TruckerCloud

Telematics aggregation and a direct line into the insurer FNOL network for instant first-notice-of-loss.

MAPS

Google 3D Tiles

Photorealistic real-world geometry so the reenactment happens on the actual street, not a diagram.

Request a demo

Put a reconstruction in front of your team.

See XShield rebuild a real collision, end to end — capture, 3D reenactment, and the court-ready dossier — on your own fleet's hardware.

Book a demo → Talk to us