RadarPub - EdgeFirst Radar Node

Real-time radar sensor processing node for the EdgeFirst Perception Middleware

RadarPub bridges automotive radar sensors to the EdgeFirst Perception stack, providing real-time target detection, clustering, tracking, and 4D radar cube publishing over Zenoh middleware. Designed for edge AI applications requiring low-latency sensor fusion on resource-constrained platforms.

• Smart Micro DRVEGRD Protocol Support - Complete CAN and Ethernet/UDP protocol implementation
• Real-Time Target Processing - Low-latency processing from CAN reception to Zenoh publish
• 4D Radar Cube Publishing - Full radar data tensor (range × azimuth × elevation × doppler)
• Advanced Clustering - DBSCAN spatial clustering for target grouping
• Multi-Object Tracking - ByteTrack algorithm with Kalman filtering for consistent track IDs
• ROS2-Compatible Output - PointCloud2 and TransformStamped message formats via Zenoh

• Seamless Zenoh Integration - Publishes to EdgeFirst Perception topics for sensor fusion
• Hardware-Optimized - Validated on Maivin and Raivin edge AI platforms
• Tracy Profiling Support - Performance instrumentation for real-time analysis
• Flexible Configuration - Runtime parameter adjustment via CLI or control utility

• Radar Sensor: Smart Micro DRVEGRD 169/174 (76-77 GHz automotive radar)
• Platforms: Maivin (NXP i.MX 8M Plus), Raivin (automotive-grade), Linux ARM64/x86_64
• Interfaces: CAN (500 kbps), Ethernet/UDP (radar cube data)

• Linux system with SocketCAN support (kernel 2.6.25+)
• CAN interface hardware (or virtual CAN for testing)
• Rust toolchain 1.70+ (for building from source)

From Binary Release:

# Download latest release for ARM64 (Maivin/Raivin)
wget https://github.com/EdgeFirstAI/radarpub/releases/latest/download/radarpub-aarch64
chmod +x radarpub-aarch64
sudo mv radarpub-aarch64 /usr/local/bin/radarpub

From Source:

git clone https://github.com/EdgeFirstAI/radarpub.git
cd radarpub
cargo build --release --features "can,zenoh"
sudo cp target/release/radarpub /usr/local/bin/

Cross-Compile for ARM64:

# Install cross-compilation tool
cargo install cross

# Build for ARM64
cross build --target aarch64-unknown-linux-gnu --release

1. Set up CAN interface (if not already configured):

sudo ip link set can0 type can bitrate 500000
sudo ip link set can0 up

2. Run RadarPub with default settings:

radarpub --can-interface can0 --zenoh-mode peer

3. View published topics:

# Using Zenoh bridge or subscriber
z_sub -t "/rt/radar/**"

RadarPub is optimized for real-time sensor processing on resource-constrained edge platforms. The system is designed to handle:

• Multiple radar targets per frame at typical sensor frame rates (10 Hz)
• Concurrent CAN and UDP data streams
• Optional DBSCAN clustering and ByteTrack tracking
• Radar cube tensor processing

Actual performance depends on hardware configuration, sensor settings, and enabled features. For deployment planning and system integration, contact [email protected].

For architecture details and tuning guidelines, see ARCHITECTURE.md.

# Basic configuration
radarpub \
  --can-interface can0 \
  --zenoh-mode peer \
  --log-level info

# Enable clustering and tracking
radarpub \
  --can-interface can0 \
  --enable-clustering \
  --cluster-epsilon 0.5 \
  --cluster-min-points 3

# Adjust radar parameters (requires drvegrdctl)
drvegrdctl --can-interface can0 set-frequency 76.5
drvegrdctl --can-interface can0 set-sensitivity high

For complete configuration options, see the User Guide.

RadarPub includes comprehensive examples demonstrating different integration patterns:

Connect directly to a radar sensor and visualize data with Rerun:

# Live radar with CAN and cube data
cargo run --example radar_viewer --features rerun -- --device can0 --cube --viewer

# Replay PCAP file for analysis
cargo run --example radar_viewer --features rerun -- radar_capture.pcap --viewer

# Record visualization to file
cargo run --example radar_viewer --features rerun -- --device can0 --record output.rrd

Use cases:

• Hardware validation and debugging
• Offline analysis of recorded data
• Direct sensor integration without middleware

Subscribe to RadarPub's Zenoh topics and visualize processed data:

# Subscribe to local RadarPub instance
cargo run --example zenoh_viewer --features rerun -- --viewer

# Connect to specific topics
cargo run --example zenoh_viewer --features rerun -- --targets --clusters --viewer

# Connect to remote Zenoh router
cargo run --example zenoh_viewer --features rerun -- \
  --zenoh-mode client \
  --zenoh-router tcp/192.168.1.100:7447 \
  --viewer

Use cases:

• EdgeFirst Perception pipeline integration
• Multi-node distributed systems
• Sensor fusion visualization

See examples/README.md for complete documentation and additional examples.

RadarPub is a core component of the EdgeFirst Perception Middleware, providing radar sensor integration for autonomous systems and robotics applications.

• EdgeFirst Perception - Multi-sensor fusion middleware

  • Combine radar data with camera, LiDAR, and IMU sensors
  • Unified coordinate frame transformations
  • Real-time sensor synchronization

• EdgeFirst Studio - MLOps Platform

  • Deploy and manage perception pipelines at scale
  • Monitor sensor health and performance
  • A/B testing and gradual rollouts
  • Free tier available for development

• EdgeFirst Modules - Hardware Platforms

  • Maivin: Edge AI development platform (NXP i.MX 8M Plus)
  • Raivin: Automotive-grade edge AI platform
  • Custom hardware design services

• 📖 User Guide: https://doc.edgefirst.ai/perception/radarpub/
• 🔧 API Reference: https://docs.rs/radarpub/
• 🏗️ Architecture: ARCHITECTURE.md (internal design documentation)
• 🚀 EdgeFirst Perception: https://doc.edgefirst.ai/perception/

• 📚 Documentation: https://doc.edgefirst.ai/perception/radarpub/
• 💬 GitHub Discussions: Ask questions and share ideas
• 🐛 Issue Tracker: Report bugs and request features

For production deployments and enterprise requirements, Au-Zone Technologies offers:

• Training & Workshops - Accelerate your team's expertise with EdgeFirst Perception
• Custom Development - Extend RadarPub or integrate additional radar sensors
• Integration Services - Seamless integration with your existing autonomy stack
• Enterprise Support - SLAs, priority fixes, and dedicated engineering support
• Hardware Services - Custom carrier boards and platform optimization

📧 Contact: [email protected] | 🌐 Learn more: au-zone.com

We welcome contributions from the community! Please see CONTRIBUTING.md for:

• Development setup and build instructions
• Code style guidelines and testing requirements
• Pull request process and review guidelines

This project follows our Code of Conduct. By participating, you agree to uphold this code.

For security vulnerabilities, please see SECURITY.md or email [email protected] with subject "Security Vulnerability - RadarPub".

We take security seriously and aim to respond to reports within 48 hours.

Licensed under the Apache License, Version 2.0. See LICENSE for details.

Copyright (c) 2025 Au-Zone Technologies. All Rights Reserved.

Third-party dependencies and attributions are documented in NOTICE.md.

• EdgeFirst Perception Team - For middleware architecture and integration support
• Smart Micro - For DRVEGRD radar protocol documentation
• Zenoh Project - For exceptional real-time middleware
• Community Contributors - See CONTRIBUTORS.md
• Open Source Projects - See NOTICE.md for complete attribution

Built with ❤️ by Au-Zone Technologies | Empowering Edge AI for Autonomous Systems