Principal Investigators: Abderazek Ben Abdallah (PI), Zhishang Wang (PI)
Our research in power/energy-efficient computing systems is critical for addressing the growing demand for more powerful and sustainable technology. As our reliance on computing devices increases, so does the need to manage their energy consumption. Efficient computing systems can reduce energy costs, minimize environmental impact, and extend the battery life of portable devices. Our research drives innovation in hardware and software design, leading to smarter, greener technologies that benefit both users and the planet.
This project introduces AIRBiS (AI-powered Reconfigurable Biomedical System), a hardware platform specifically engineered for scaling deep-learning inference in resource-constrained environments. By employing specialized hardware accelerators, the platform can perform complex medical diagnostics with an accuracy of 95.2%. The architecture is designed to be self-contained and reconfigurable, adaptive to various neural network topologies while maintaining an energy-efficiency profile 13 times better than conventional general-purpose processors.
Traditional Artificial Neural Networks (ANNs) are often computationally expensive and ill-suited for real-time edge devices. Our research focuses on Spiking Neural Networks (SNNs) integrated into 3D Network-on-Chip (3D-NoC) architectures. By mimicking the brain's spatio-temporal information processing, these neuromorphic systems offer high reliability and fault tolerance. Our proposed 3D-NoC routing scheme ensures robust communication even under hardware failure, consuming 32% less power than standard ANN-based systems.
This patented technology focuses on a novel AI Processor architecture designed for high-throughput and energy-efficient execution of modern machine learning workloads. The design emphasizes a modular hardware approach that optimizes data flow between processing elements, significantly reducing the energy overhead associated with memory access. This architecture serves as the foundation for the next generation of embedded AI systems, providing necessary computational power for autonomous edge devices.
[Patent No. 7699791] (June 20, 2025) ? Abderazek Ben Abdallah, Hoang Huang Kun, Dang Nam Khanh, Song Janning, "AI Processor" [Source]
As healthcare shifts toward personalized and remote monitoring, there is a critical need for SoCs that can handle real-time signal processing on-body. This project explores the design of embedded multicore systems specifically for ECG (Electrocardiogram) processing. Unlike traditional systems that merely collect and transmit data, our SoC performs local, real-time feature extraction and anomaly detection, ensuring long-term use on portable and wearable devices.
The Queue computation model provides an alternative to traditional register-based architectures by utilizing a circular queue-register to manage operands. This project researches a novel parallel processor that eliminates "false dependencies," which are a primary source of complexity and power consumption in modern CPUs. By removing the need for power-hungry hardware like register renaming, the Queue Processor achieves high instruction-level parallelism (ILP) with minimal hardware overhead. Our research includes the development of a complete cross-layer tool-chain?including ILP-aware compilers and cycle-accurate simulators?to optimize the execution of embedded applications.
