[AUTO] Automotive & Mobility

Edge AI and telematics for vehicles — where OTA touches functional safety.

OTA in a consumer IoT device that fails badly means an unhappy customer. OTA in a vehicle system touches functional safety, and that changes the update pipeline from top to bottom. Rivix builds automotive-grade firmware and the fleet-facing platform around it as one engagement.

What this covers

The full stack, scoped for automotive-grade delivery.

Technical considerations

What actually shapes the architecture in this space.

OTA that has to be provably safe, not just reversible

The consumer OTA fundamentals — dual-bank (A/B) partitioning, signature verification, staged rollout — carry over, but rollback has to be provably safe and tied to the vehicle's own safety-relevant self-checks, not a generic health check. Automotive OTA manages update sets — coordinated bundles of firmware versions across multiple ECUs validated to work together — with rollback needing the same coordination in reverse. Any OTA to an ECU under functional-safety process can't ship until formal safety validation and sign-off completes, a slower gate than typical consumer OTA cadence.

ISO 26262 and ASIL classification

ISO 26262 is the international functional safety standard for electrical and electronic systems in road vehicles. Components are classified ASIL A through D (plus QM) based on a severity/exposure/controllability hazard analysis, so a braking-related component carries different rigor than an infotainment feature. It requires forward traceability from requirement through design to test, maintained continuously through a structured V-model development process.

Telematics sensor fusion

GPS gets fused with vehicle speed from the OBD-II bus or CAN, plus inertial data from an onboard accelerometer/gyroscope, with dead-reckoning to bridge GPS gaps in tunnels, urban canyons and parking structures. Position typically reports on a distance/time threshold rather than a fixed high-frequency interval, and derived events — harsh braking, idling, geofence entry/exit — get computed on-device and sent instead of a raw data stream.

Why the dash unit runs inference at the edge

Edge inference was chosen over cloud inference specifically to hit sub-200ms alerting latency and to avoid the bandwidth cost and privacy exposure of continuously streaming raw video from every vehicle in a fleet. Only classified events and short clips are transmitted, with local buffering carrying events through connectivity gaps.

Proven in the field

Real builds in automotive and mobility.

Pulled from our case-study library — same AI/human split shown in full, industry context included.

Driver-Behavior Edge-AI Dash Unit & Alert System

US-based fleet-management / automotive technology company

An in-vehicle edge-AI system analyzes driving behavior locally and detects defined safety events in real time without requiring continuous transmission of raw sensor/video data.

~38%Estimated AI-first engineering effort reduction
Read full case study →

EV Battery-Pack Thermal Monitoring Module

US-based EV / battery technology company

A distributed monitoring module captures thermal conditions across an EV battery pack and identifies abnormal temperature behavior for battery-management and diagnostic systems.

~28%Estimated AI-first engineering effort reduction
Read full case study →
Field Notes

What we've written about building in this space.

[AUTO]

Over-the-Air Updates for Automotive-Grade Embedded Systems

OTA that touches functional safety changes the update pipeline from top to bottom, not just the risk tolerance.

Read the article →
[AUTO]

Designing Telematics Systems for Fleet and Mobility Applications

Serving a dispatcher who needs real-time location and an analyst who needs months of history, from the same data.

Read the article →
[AUTO]

Functional Safety Basics: What ISO 26262 Means for Embedded Teams

ISO 26262 isn't a checklist bolted onto an existing process — it's a different process.

Read the article →
Common questions

What people ask before scoping an automotive product.

How do you handle OTA safely in a vehicle system?

Signed, rollback-capable OTA pipelines with staged rollout — the update path is designed alongside the firmware, not bolted on after v1 ships.

Are you set up for ISO 26262 work?

Teams moving into automotive quickly discover ISO 26262 isn't a checklist bolted onto an existing process — it's a different process, and we scope engagements with that gap in mind.

Can you process raw sensor or video data on-device instead of streaming continuously?

Yes — our edge-AI dash unit work detects defined safety events locally, sending only the event, not continuous raw video.

What's the AI/human split in practice here?

AI scaffolds drivers and protocol-stack boilerplate; a senior firmware engineer owns real-time timing guarantees and every power-critical code path.

Building automotive-grade firmware or a fleet telematics platform?