Energy-efficient in-memory computing with 8T SRAM for arithmetic operations and signal filtering in UAV communications

Abstract

The increasing demand for data-intensive artificial intelligence and machine learning applications has exposed the limitations of traditional Von Neumann architectures, especially in resource-constrained environments like Unmanned Aerial Vehicle (UAV) communication systems. This work introduces an advanced in-memory computing model leveraging an 8T SRAM-based architecture combined with a multi-logic sense amplifier to perform arithmetic operations directly within the memory array. By embedding processing into the memory, this approach significantly reduces data transfer overhead, resulting in lower latency and improved energy efficiency – key requirements for UAV systems. Additionally, a novel lightweight and energy-efficient signal processing method is proposed. This architecture enables real-time signal filtering, effectively minimizing noise and enhancing signal integrity while meeting the compactness and scalability demands of UAV systems. Simulation results demonstrate significant reductions in power consumption and latency across a range of arithmetic operations, with robust performance maintained under varying process, voltage, and temperature conditions. This transformative design offers a practical and efficient solution for next-generation aerial communication technologies, ensuring high-quality communication and efficient data processing in critical UAV applications.