NEUROPULS: NEUROmorphic energy-efficient secure accelerators based on Phase change materials aUgmented siLicon photonicS

Abstract

The Horizon Europe NEUROPULS project introduces novel secure and energy-efficient neuromorphic accelerators leveraging Phase Change Material (PCM) augmented silicon photonics technology. These accelerators are designed to interface seamlessly with RISC-V architectures via an FPGA-powered computing platform. The research targets critical edge-computing applications, including GNSS anti-jamming, autonomous driving, and anomaly detection, with a strong emphasis on security and reliability.

Report

Key Highlights

• Project Goal: Development of secure and ultra-energy-efficient neuromorphic accelerators for edge computing under the Horizon Europe NEUROPULS project.
• Core Innovation: Utilization of augmented silicon photonics technology, leveraging Phase Change Materials (PCMs) for neuromorphic functions.
• Interface Standard: Designed to be interfaced directly with RISC-V architectures.
• Development Platforms: The project aims to deliver three core components: an augmented silicon photonics platform, an FPGA-powered RISC-V-connected computing platform, and a comprehensive simulation platform.
• Target Use Cases: Edge applications requiring high performance and security, including Global National Satellite System (GNSS) anti-jamming, autonomous driving, and anomaly detection.

Technical Details

• Architectural Stack: The accelerators rely on integrating non-volatile and adjustable components (PCMs) directly within silicon photonic waveguides to perform critical neuromorphic operations (e.g., matrix-vector multiplication).
• Integration Method: RISC-V serves as the central processing unit, utilizing an FPGA-powered platform to efficiently connect and manage the high-speed photonic accelerator, enabling real-time edge computation.
• Evaluation: The project involves detailed analysis using a simulation platform to address the main advantages and limitations of the underlying augmented silicon photonics technology.
• Non-Functional Requirements: A significant focus is placed on assessing and ensuring the reliability and security of the stand-alone accelerator implementation, especially crucial for safety-critical applications like autonomous systems.

Implications

• Advancing RISC-V Heterogeneity: This work strongly validates the RISC-V ecosystem as the ideal open foundation for integrating disruptive, next-generation accelerators like photonic neuromorphic processors, expanding RISC-V’s role in high-performance edge AI.
• Breakthrough in Energy Efficiency: By merging neuromorphic principles with silicon photonics, NEUROPULS promises orders of magnitude improvements in energy efficiency compared to traditional electronic hardware, which is vital for sustained, battery-powered edge devices.
• Enabling Critical Edge AI: The targeted use cases (GNSS anti-jamming, autonomous driving) require ultra-low latency and verifiable security. The project’s emphasis on reliability and security helps transition optical computing from research to mission-critical, real-world deployment.
• Scalable Photonic Memory: Utilizing Phase Change Materials addresses the challenge of building dense, non-volatile, and programmable weight storage necessary for scalable, high-density optical neural networks.