A Cloud-Aware Scalable Architecture for Distributed Edge-Enabled BCI Biosensor System

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Existing Brain-Computer Interface (BCI) biosensors struggle with scalability, latency, and cloud integration.
• Continuous neural activity monitoring requires robust and efficient data processing frameworks.

Purpose of the Study:

• To develop and validate a cloud-aware, real-time cognitive grid architecture for multimodal BCI biosensors.
• To integrate edge-resident TinyML (Tiny Machine Learning) with cloud infrastructure for enhanced BCI data analytics.

Main Methods:

• A physical prototype integrating BioAmp EXG Pill, RP2040 microcontroller, and environmental sensors was developed.
• A tiered data pipeline (SD card, Redis, PostgreSQL, AWS S3) managed data flow from acquisition to archival.
• Edge-resident TinyML enabled on-device feature extraction and inference, communicating via Wi-Fi/4G.

Main Results:

• The system achieved consistent edge-level inference with bounded latency and packet loss below 5%.
• Cloud-assisted analytics showed variable delays typical of cellular networks and serverless computing.
• Hybrid deployment strategies facilitated cost-efficient validation while maintaining architectural integrity.

Conclusions:

• The developed framework is scalable, modular, and energy-efficient, supporting advanced analytics for translational BCI applications.
• The architecture effectively combines edge-resident TinyML inference with cloud-based machine learning workflows.
• This work lays the foundation for future BCI research and development with improved data handling capabilities.