Share

Make radar smarter

#AI #Graph Neural Networks #RaGNNaR
Credit: Generated with AI

Overview

Graph Neural Networks (GNNs) are a powerful emerging AI model designed to process data structured like graphs. They excel at identifying complex patterns and relationships within data, making them ideal for handling the irregular data produced by sparse sensors like radar, LiDAR, and event cameras.

Our RaGNNaR demo demonstrates how GNNs effectively process radar data for gesture recognition using standard off-the-shelf sensors. By leveraging unique data features like Doppler and intensity, GNNs perform classification tasks with remarkable efficiency and minimal computational effort.
CEA offers complete expertise in integrating GNN solutions. We provide tailored approaches from nearsensor processing to edge devices, addressing the specific challenges of sparse sensor data.

 

Credit: Generated with AI

Benefits

  • Key-enabler to integrate near-sensor computing
  • Reduced bandwidth (local processing)
  • Real-time processing (fast feedback control)
  • Enhanced security (raw data contained within the sensor)
  • Frugal solutions: low computing, power and memory footprint
  • Low cost

 

Key metrics vs others

Applications >

Mobility (Automotive)

In-Cabin sensing: Advanced driver and passenger safety systems ; Out-Cabin sensing: Enhanced vehicle autonomy and safety

Applications>

Industrial & Defense

Drone integration: Monitoring and inspection

Applications>

Smart City & Smart Home

Gesture control: Intuitive human-machine interfaces ; Sleep analysis: Advanced monitoring systems

Applications>

Healthcare

Fall detection: Safety systems for vulnerable individuals ; Non-Invasive diagnostics: Vital sign monitoring solutions

Trends

  • Real-time
  • Low power
  • Data privacy
  • Secure

End-to-end offer

Proven Expertise in coupling data, algo and HW:

  • Algorithm optimization: Multimodal sensor data processing
  • AI techniques: Cutting-edge approaches for maximum performance
  • Implementation solutions: Co-design and dedicated architectures

 

Succes story:

  • CEA has successfully implemented GNN solutions with event cameras and developed integrated chip solutions

 

We target multi-sensor solutions:

  • Heterogeneous sensors: LiDAR, Camera, Radar and much more.

Contact

To go further