Software Architecture

Our autonomous system operates on Ubuntu 20.04 with ROS2 Foxy. Our pipeline uses a distributed node-based approach where individual components communicate through the ROS2 middleware, enabling scalable and maintainable system design.

System Overview

Node Architecture

Our software follows the standard ROS2 publisher-subscriber communication pattern:

Publishers

Sensor nodes publish raw data to specific topics
Processing nodes publish filtered/processed data
Control nodes publish command messages

Subscribers

Processing nodes subscribe to sensor data topics
Decision-making nodes subscribe to processed data

This decoupled architecture allows nodes to be developed, tested, and deployed independently while maintaining system-wide coordination.

Startup Procedure

The system launches automatically when the vehicle’s low voltage (LV) system powers the compute unit:

Power-On: LV system activates, powering the compute
Boot Sequence: Ubuntu 20.04 loads and initializes
Cron Launch: System startup triggers cron job on @reboot
Tmux Session: Main software pipeline launches in a tmux session
Node Initialization: All ROS2 nodes start in predetermined order
System Ready: Heartbeat monitoring begins once all nodes are active

The cron job ensures the system launches reliably without manual intervention, while tmux provides session persistence and easy debugging access.

Heartbeat System

Since any node failure during autonomous driving is run-ending, our heartbeat architecture focuses solely on ensuring reliable system state prior to ready-to-drive (RTD):

Monitoring Cycle

Heartbeat pulses every 5 seconds
Checks for node presence in ROS2 graph
Verifies active publishing on expected topics
Restarts failed nodes automatically

Pseudocode Reference

HEARTBEAT MONITOR:

REPEAT every 5 seconds:
    FOR each critical node:
        IF node does not exist in ROS2 graph:
            Log error message
            Restart the node
            Continue to next node

        FOR each expected topic from this node:
            IF topic is not publishing data:
                Log error message
                Restart the node
                Break out of topic loop

    Log "All nodes healthy" message


RESTART NODE PROCEDURE:
    Log restart message
    Kill the node process
    Wait 2 seconds for cleanup
    Launch the node again


CHECK NODE EXISTS:
    Get list of all active nodes
    Return true if node name is in the list


CHECK TOPIC PUBLISHING:
    Echo topic to see if it has received messages recently
    Return true if message rate is greater than 0

This monitoring system ensures a reliable boot prior to RTD, allowing for safe and reliable remote startup.

Software Architecture

System Overview

Node Architecture

Startup Procedure

Heartbeat System

More CMR Driverless