DeepSig @Mobile World Congress 2024
Mobile World Congress 2024 witnessed a landmark moment in the realm of wireless communications as DeepSig unveiled breakthrough innovations centered on the integration of deep learning (DL) into the signal processing at the heart of 5G radio access networks (RAN). Representing the culmination of nearly a decade of pioneering research and development, these advancements are reshaping the wireless landscape and pave the way for a fully AI native next generation air interface. In collaboration with industry leaders, DeepSig showcased the power of artificial intelligence (AI) to enhance network efficiency, optimize spectrum utilization, and revolutionize air interface design.
The question I heard most often at the show was why DeepSig was mentioned alongside the largest tech companies in the conversation about how AI is transforming wireless infrastructure. This MWC recap should help answer that question. But first, some context: The industry’s earliest inventions in this field date back to concepts put forth in 2015 by our co-founder and CTO, Tim O’Shea, during his tenure as a research professor at Virginia Tech. Some examples are contained his seminal publication, co-authored with Jakob Hoydis, “An Introduction to Deep Learning for the Physical Layer.” The paper explores how deep learning (DL) can revolutionize wireless communications with a new and innovative way of optimizing transmitter and receiver components together, akin to automating the reconstruction of a puzzle.
This novel approach, including channel autoencoders (CAEs) as well as radio transformer networks (RTNs), incorporates wireless expert knowledge into DL models to enhance system performance with the best of both domains. Additionally, the paper showcases how convolutional neural networks (CNNs) improve modulation classification and herald a new era in communications design that promise significant advances in tackling some of the most complex and intricate wireless communications challenges in ways traditional human constructs cannot. Tim’s early patents at Virginia Tech followed him to DeepSig, the company he co-founded with Jim Shea, CEO, in 2016. Since then, DeepSig has amassed 100+ patents[1] and commercialized two key software product lines: OmniPHY (3GPP layer 1 processing) for public and private 5G markets, and earlier, OmniSIG (RF sensing/classification), for tactical/defense applications.
DeepSig’s accelerating momentum at MWC24:
_______________________________________________________________________________________
World’s First AI Native Neural Receiver in a Live 5G Macro Network
The collaboration between Viettel High Tech (VHT) and DeepSig resulted in the deployment of the industry’s first live AI/ML-based neural receiver in a 5G Open virtualized radio access network (vRAN) across Viettel’s network in Hanoi.
DeepSig’s OmniPHY-5G neural receiver software, integrated into VHT’s Distributed Units (DU), leverages AI/ML algorithms to enhance uplink throughput and resource utilization. By improving channel estimation and equalization in the physical layer, the solution improves the Signal to Interference and Noise Ratio (SINR) of the shared uplink channel, thereby boosting 5G RAN capacity by a factor of 2-3X without additional hardware. Read the Press Release
AI-RAN Alliance Co-Founded by DeepSig and Other Industry Leaders
DeepSig joined forces with leading technology companies, industry players, and academic institutions to launch the AI-RAN Alliance. This collaborative initiative, which includes top AI compute, cloud and telecom infrastructure companies, aims to integrate AI into cellular technology, advancing RAN capabilities, and driving innovation in mobile networks.
Through collective research efforts, the alliance focuses on three main areas: AI for RAN, AI and RAN integration, and AI on RAN. By harnessing AI, the alliance seeks to enhance network efficiency, reduce power consumption, and unlock new revenue opportunities for telecommunications companies. Read the Press Release
Commercial Release of Industry-First Neural Receiver AI Software
DeepSig announced the commercial release of its Gen 1 OmniPHY 5G software, marking the industry’s first 5G carrier-grade neural receiver. The software, integrated with Intel FlexRAN Layer 1 (L1) reference software, revolutionizes the physical layer processing in Open vRAN 5G networks.
By employing deep neural networks for channel estimation, equalization, and interference rejection, OmniPHY 5G delivers a 2-3X increase in uplink throughput, superior spectral efficiency, and lower total cost of ownership (TCO). The release sets the stage for AI-native 6G technology, paving the way for transformative innovations in wireless communications. Read the Press Release
Partner Activities:
Intel AI Development Kit with DeepSig
Building on our ongoing collaboration, which incorporated DeepSig’s industry-first AI-native L1 software into FlexRAN, Intel announced at MWC an AI Development Kit for vRAN to accelerate the deployment of AI-driven solutions in 5G radio networks. Read the Press Release
Anritsu and DeepSig Enable RF Awareness with AI
At MWC, Anritsu demonstrated how it is revolutionizing spectrum sensing by integrating DeepSig’s AI machine learning technology with its Field Master Pro spectrum analyzer, enabling real-time signal detection and classification that adapts to changing RF conditions and new types of emitters allowing for powerful real time monitoring and optimization of spectrum utilization. Read the Press Release
DeepSig’s growing industry partnerships underscore the industry-wide recognition of our software and expertise in AI-driven wireless communications.
AI Native Next Generation Air Interface
DeepSig showcased a pre-6G modem demo at MWC 2024 that offers a glimpse into the future of deep learning in wireless communications. Fundamentally, this new approach to air interface design and operation leverages AI/ML to learn new waveforms and modulation on the fly at both the transmitter and receiver and to take RAN performance, efficiency and capabilities to the next level.
DeepSig has deployed these techniques for tactical/defense applications and in its ongoing next-gen R&D, laying groundwork for emerging 6G standards. See the Demo
Conclusion
The announcements at MWC 2024 underscore our commitment to driving innovation in wireless communications through AI, a journey started by our founders nearly a decade ago. By harnessing the power of deep learning, DeepSig is redefining the capabilities of the RAN and laying the foundation for the future of wireless technology. With its groundbreaking technology and solutions, DeepSig is uniquely positioned as the global technology leader in AI-native wireless communications, helping its customers and partners unlock new possibilities for network efficiency, spectrum utilization, and end-to-end performance optimization.
Written by: Peter Carson, Director, AI/ML Product Management
About DeepSig
DeepSig is a pioneer in AI-native wireless communications. Its transformative technology pushes the boundaries of wireless performance, spectrum visibility and network capabilities. Drawing on a unique blend of artificial intelligence, deep learning and wireless systems expertise, DeepSig’s AI/ML powered software unlocks unprecedented gains in capacity, efficiency and security for specialized communications networks and 5G Open RAN, laying the foundation for AI-native 6G.
[1] Issued or pending.
