Millimeter wave wireless network on chip using deep reinforcement learning

Abstract

Wireless Network-on-Chip (NoC) has emerged as a promising solution to scale chip multi-core processors to hundreds of cores. However, traditional medium access protocols fall short here since the traffic patterns on wireless NoCs tend to be very dynamic and can change drastically across different cores, different time intervals and different applications. In this work, we present NeuMAC, a unified approach that combines networking, architecture and AI to generate highly adaptive medium access protocols that can learn and optimize for the structure, correlations and statistics of the traffic patterns on the NoC. Our results show that NeuMAC can quickly adapt to NoC traffic to provide significant gains in terms of latency and overall execution time, improving the execution time by up to 1.69× -3.74×. CCS CONCEPTS• Networks → Network protocols; Wireless access networks.

Electromagnetic Nanonetworks Beyond 6G: From Wearable and Implantable Networks to On-Chip and Quantum Communication

“…In particular, a cross-disciplinary approach where the architecture guides the MAC protocol operation, instead of the protocol blindly trying to adapt to the traffic, is promising and unique to this monolithic scenario. This is the strategy followed in [135], where a MAC protocol was designed to give priority to latency-sensitive messages from the computer architecture; or like in [136], where protocols were simplified (and made faster) exploiting the fact that certain compute applications are tolerant to communication errors.…”

“…More examples of cross-layer design could be found in application-specific scenarios. For instance, in the context of wireless chip-scale networks, MAC protocols could benefit from application-level information such as message criticality, which can then be used to drop packets early to reduce temporary congestion without affecting the program's output [136] or to prioritize transmissions that optimize application metrics (execution time) rather than network metrics (communication latency) [135]. Further, the wireless channel within package can be used as a medium for collective computation through the overlap of concurrent transmissions, to implement thread synchronization primitives [186] or majority gates [187] useful for certain parallel computing architectures.…”

Metasurface‐Programmable Wireless Network‐On‐Chip

“…This is expected in a rich-scattering system and in contrast to the (analytically calculated) RIS patterns used for beamforming in free space. [8] To illustrate the generality of our technique, as well as its realtime in situ adaptability to dynamic traffic patterns, [80] we apply it also to two other pairs of antenna locations on a chip with a 100 μm thick silicon layer as well as to one case on a chip with a 200 μm thick silicon layer. The effectiveness of our technique is clearly seen for all three cases in Figure 5.…”

Channel Estimation for Reconfigurable Intelligent Surface-Aided mmWave Communications

“…To demonstrate the 3D beamforming capability, we fabricate a 4x4 metasurface whose unit-cells are separately controlled. 8 Related Works mmWave communication is considered to be the future of mobile communications, on which extensive research has been conducted [16,27,33,38,54]. However, mmWave communication suffers from high attenuation, such that antenna arrays for effective beamforming are necessary [20,32,35,36,39,50,53].…”