We built Co-Driver AI Middleware – a sophisticated real-time vehicle telemetry system that processes CAN bus and OBD-II data to provide AI-powered contextual alerts and driver assistance. The system reads 1000+ CAN messages per second, processes complex alert rules with <200ms latency, and provides intelligent telemetry summaries for AI reasoning.
/ 01 — THE CHALLENGE
The Problem
High-performance vehicle owners and racing teams face a critical problem: how to make sense of hundreds of real-time telemetry signals while driving. Traditional dashboards show data, but they don’t provide context, alerts, or intelligent assistance.
In racing or high-performance driving, milliseconds matter. A warning light isn’t enough, drivers need to know what it means, how urgent it is, and what action to take. Push these cars too hard without monitoring critical systems and you’re looking at an expensive bill: a broken transmission or blown engine.
/ 02 — INSPIRATION
The Inspiration
As a track day enthusiast, there’s a constant balancing act: one eye on the track, the other on the gauges. The idea behind Co-Driver is simple but powerful: have AI monitor the CAN bus and issue voice alerts when something is about to go wrong. It’s about pre-empting failure before it happens.
/ 03 — THE SYSTEM
The Solution
We built Co-Driver AI Middleware, a sophisticated real-time vehicle telemetry system that processes CAN bus and OBD-II data to provide AI-powered contextual alerts and driver assistance. The system reads 1000+ CAN messages per second, processes complex alert rules with less than 200ms latency, and provides intelligent telemetry summaries for AI reasoning.
Real-time Capabilities
/ 04 — VALIDATION
Test Vehicles
We validated the system on two Nissan GT-R platforms with different ECU configurations. The R33 uses a Syvecs S8 with direct CAN bus integration, while the R35 runs an EcuTek Stage 4.25 tune with OBD-II Bluetooth. Both architectures are supported.
Nissan Skyline R33 GT-R
Syvecs S8 ECU with direct CAN bus integration. High-frequency telemetry capture for track-day monitoring.
Nissan R35 GT-R
EcuTek Stage 4.25 with OBD-II Bluetooth integration. Universal compatibility through standard diagnostic port.
/ 05 — CAPABILITIES
What It Does
/ 06 — WORKFLOW
How It Works
The pipeline runs through four stages: ingest raw data from the vehicle, map signals to a universal format, evaluate alert rules in real-time, and dispatch alerts or context updates to the AI assistant. Each stage is optimised for minimal latency.
1
Ingest
CAN Bus or OBD-II captures raw data from the vehicle ECU at high frequency.
2
Map
Signal Mapper translates raw signals to a universal format regardless of vehicle type.
3
Evaluate
Alert Engine processes multi-condition rules with logical operators in under 10 milliseconds.
4
Act
Alert Dispatcher sends immediate alerts or context updates to the AI assistant for voice guidance.
/ 07 — FEATURES
Key Features
The system combines high-frequency data processing, dual-source input, universal signal mapping, and an enhanced alert engine. Here’s what makes it work at racing speeds.
/ 08 — ALERTING
Alert System and Drive Modes
The alert system works on two tiers: immediate API interruptions for critical conditions like oil pressure failure, and context updates every 30 seconds for ongoing AI awareness. The system also supports different drive modes with conservative thresholds for street driving and performance thresholds for the track.
Immediate Alerts
Direct API interruption for critical conditions like oil pressure failure or coolant temperature spikes. Millisecond response time for driver safety.
Context Updates
30-second telemetry injection for ongoing AI awareness. Intelligent batching reduces noise while keeping the assistant informed of vehicle state.
Drive Modes
Street Mode with conservative thresholds for daily driving, and Track Mode with performance thresholds for racing. Dynamic real-time switching with visual confirmation.
/ 09 — ARCHITECTURE
Technical Architecture
For the technically inclined, here’s what’s under the hood. Python services handle CAN bus acquisition, OBD-II protocol, signal processing, and the enhanced alert engine. Plotly Dash provides the real-time frontend. The CodriverAI API exposes a two-tier alert system with context injection endpoints.
Backend
Python services for CAN bus acquisition, OBD-II protocol, signal processing, universal mapping, and enhanced alert engine with ENUM support.
Frontend
Plotly Dash for real-time telemetry visualisation with live gauges, charts, and a visual rule builder interface.
Integration
CodriverAI API with two-tier alert system and context injection endpoints for voice-based driver assistance. 400+ PID database for accurate interpretation.
/ 10 — OUTCOME
The Results
We built this because we needed it for our own track days. The system now runs on both test vehicles, processing thousands of CAN messages per second with voice alerts that warn drivers before things go wrong. The live dashboard lets anyone monitor vehicle state in real-time, and the alert engine catches issues that human eyes would miss.
Want AI-Powered Vehicle Intelligence?
Whether you're building AI for vehicles, industrial equipment, IoT devices, or any real-time data source, we figure out where AI can help and build bespoke solutions.
Start Your Project