6GEM+

Within 6GEM+, the Chair of Embedded Systems, led by Prof. Akash Kumar at Ruhr University Bochum, develops energy-efficient, high-performance, and reconfigurable computing architectures for latency-sensitive AI-native 6G applications. The research investigates application-specific precision optimization, approximate arithmetic operators, and automated design-space exploration to develop resource-efficient 6G transceiver accelerators, particularly on FPGAs, balancing accuracy, throughput, hardware resources, and energy consumption. For distributed AI processing, the project explores in-network computing, adaptive workload partitioning, and execution across heterogeneous edge devices. Various techniques including quantization, pruning, and early-exit neural networks aim to reduce computation, communication overhead, and end-to-end latency.

• Dr.-Ing. Zahra Ebrahimi
• M.Sc. David Silbermann

• Z. Ebrahimi, M. Eslami, A. Kumar, “X-DINC: Toward Cross-Layer ApproXimation for the Distributed and In-Network ACceleration of Multi-Kernel Applications”, Future Generation Computer Systems 172, 107864.
• D. Silbermann, Y. Liu, A. Kumar, “Mixed-Precision Logarithmic Quantization for Efficient Neural Network Inference on FPGAs”, IEEE Embedded Systems Letters.