A robust and scalable neuromorphic communication system by combining synaptic time multiplexing and MIMO-OFDM

Area of Science:

• Neuromorphic Engineering
• Communication Systems
• Computer Architecture

Background:

• Current neuromorphic systems face challenges in robust and efficient communication, particularly for large-scale implementations.
• Existing architectures often lack flexibility in connectivity and scalability, hindering widespread adoption.
• Need for communication methods that are both power-efficient and capable of high-speed data transfer.

Purpose of the Study:

• To propose and analyze a novel neuromorphic communication architecture.
• To enhance communication robustness, efficiency, and scalability in spiking neural networks.
• To provide a power and area-efficient solution for hardware implementation of very large-scale neuromorphic systems.

Main Methods:

• Integration of synaptic time multiplexing (STM) for intragroup communication, offering firing rate independence and flexible connectivity.
• Application of wired multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) for efficient intergroup communication.
• Simulation of the proposed architecture using large-scale spiking neural network models.

Main Results:

• The proposed neuromorphic system with MIMO-OFDM demonstrates robust and efficient communication.
• The system achieves real-time operation with a high bit rate.
• Combined STM and MIMO-OFDM techniques result in flexible, scalable connectivity.

Conclusions:

• The novel architecture effectively addresses communication limitations in neuromorphic systems.
• The integration of STM and MIMO-OFDM provides a significant advancement for large-scale hardware implementations.
• This solution offers a power and area-efficient pathway for future neuromorphic computing.