DeepSig and Software Radio Systems
in collaboration with the US DOW Future G Office
Abstract
The Open Centralized Unit Distributed Unit (OCUDU) initiative accelerates wireless innovation by creating a truly multi-vendor, open and fast-moving software-driven ecosystem. Through collaboration between the Department of War’s Future G Office, DeepSig, and Software Radio Systems (SRS), OCUDU transforms 5G/6G development from an exclusive domain of large vendors to an open playground for innovators everywhere. By combining production-ready, carrier-grade, open-source CU/DU software that operates across multiple vendors (e.g., Intel/AMD x86, ARM, NVIDIA, and other architectures), OCUDU enables anyone to develop custom network solutions on their preferred hardware, delivering commercial solutions that are more interoperable and customizable at the open-source software level. OCUDU leverages the proven SRS platform and incorporates additional features from the ATIS Minimum Viable Profile (MVP) profile, alongside accelerated compute to deliver scalable processing performance and efficiency. This robust foundation enables the deployment of gNBs that are compatible with any Open Fronthaul RU and O-RAN-compliant ecosystem, supporting seamless integration and delivering competitive performance and scalability. It also lays the groundwork for integrating artificial intelligence into the RAN (AI-RAN), referring to the application of AI and ML techniques directly with the RAN stack to optimize functions such as spectrum use, scheduling, and interference mitigation, and other advanced capabilities, allowing public and private 5G and NextG networks to be tailored for a wide range of use cases without the need to rebuild the entire stack. This approach is expected to help accelerate industry innovation, pre-standardization activities, and broad adoption, thereby improving US competitiveness and the pace of the telecom market.
The Paradigm Shift: Accelerating Standards Organizations with Open Source
The telecommunications landscape faces a fundamental challenge: standards bodies, such as 3GPP, operate in multi-year cycles, while innovation moves at software speeds. The OCUDU initiative introduces a new model: win the 6G race through open source velocity and implementation, not just in formal committee consensus.
OCUDU delivers fully open source CU/DU implementations—the core of any gNB deployment and the core of the forthcoming 6GnB—running seamlessly across Intel x86, AMD, ARM, NVIDIA, and Ampere processors, among others. Governed under vendor-neutral Linux Foundation and OSI best practices and permissive open-source licensing models, OCUDU combines technical rigor and competitive performance with openness. This flexibility transforms deployment economics and accelerates the pace of innovation. Organizations can leverage existing OpenRAN and OCUDU infrastructure while contributing to and building on a platform that evolves in days and weeks, not years of study and deliberation.
The OCUDU platform’s O-RAN Alliance compliance ensures ecosystem compatibility through well-defined, interoperable interfaces, including O-FH, F1, O1, E1, E2, and A1. But more importantly, OCUDU enables innovation beyond what standards envision. When researchers develop new AI algorithms for RAN, they can deploy and test on real networks within days. When the U.S. government identifies strategic opportunities or unique requirements, these capabilities can be integrated and deployed at software speed. This velocity advantage—powered by open-source collaboration and a vetted common open foundation—ensures rapid technological advancement and ecosystem growth.
Building on Proven Software and Ecosystem
SRS brings a decade of proven software excellence to OCUDU. Their srsRAN platform powers production networks globally, spanning commercial operators, private networks and research institutions. This mature software stack and ecosystem provide the stable foundation on which OCUDU will be built, enabling rapid innovation and the deployment of mature carrier-grade capabilities in a high-quality vendor-neutral code base. As part of the OCUDU effort, the new LF OCUDU project will build upon this and rapidly integrate new features, including multiple ATIS MVP profiles, 5G NTN, RedCap, accelerated compute features and more.
SRS’s proven software architecture and code quality mean developers can immediately focus on differentiation rather than reimplementation. When a cybersecurity company adds AI-powered threat detection to their private 5G network, they can build, integrate, and test on thoroughly tested protocol stacks – and rapidly bring a new capability to market. When manufacturers implement low-latency controls or indoor factory positioning, they leverage optimized processing and schedulers refined through years of real-world deployment, along with well-tested OpenRAN interfaces and device integration. This combination of openness and reliability accelerates time-to-market and enables the development of more diverse and innovative solutions that drive competitiveness.
DeepSig’s Central Role: Accelerating AI-RAN on OCUDU
DeepSig is delivering accelerated compute offload for key L1 processing functions to enable improved scaling and energy efficiency, while also developing key AI-RAN reference designs for OCUDU, helping to transform and solidify theoretical concepts and use cases into deployable systems and extensible open software modules that can be used for development, study, and comparison of capabilities between multiple vendors. Building upon SRS’s heavily vetted 5G stack, we leverage our experience in developing and deploying AI-RAN solutions such as OmniSIG and OmniPHY-5G to make it easy to integrate and measure the performance of similar capabilities, including AI-native processing, spectrum sensing, massive MIMO optimization, beamforming, and additional edge AI inference.
Our efforts will help design and build AI-RAN infrastructure and tools across OCUDU. We’re creating open reference designs, stack integration, dataset collection tools, and rigorous reproducible benchmarks that define how intelligence integrates with future networks for key use cases—such as resilient neural receivers, RAN digital twins, AI-native air interface prototypes, spectrum sensing and active-ISAC capabilities that transform base stations into multi-function sensors, and AI/ML-improvements to other critical data-driven functions in the RAN stack. These efforts help to lay out open reference designs which can be extended and customized by a wide range of vendors to integrate capabilities with the stack to meet a broad range of mission requirements, enabling future networks to operate with greater intelligence and adaptability
Adoption of key AI features alongside RAN functions offers to deliver significant new capability and performance but can be complex, touching data and functions across multiple layer of the stack, often and requiring integration with numerous data elements, KPIs, and non-standard controls and feedback mechanism. This presents challenges for traditional standardization processes, while open source is uniquely poised to help arrive at reference designs which help in attaining reproducibility, uniformity, and comparison across multiple-vendors. DeepSig’s AI-RAN reference designs help to enable OCUDU ecosystem members to achieve these functions rapidly, in days or weeks where standards-only approaches may have taken years, rapidly translating industry AI-RAN concepts into foundational baselines to accelerate the ecosystem. We hope this will bolster faster onramps for innovation and support broader work within the AI-RAN Alliance, Next-G Alliance, OpenRAN Alliance, 3GPP, TIP, industry, academia, and government. This isn’t incremental progress—it’s making sure we can rapidly build a foundation for 6G and AI-native wireless networks that can be tested, compared, iterated-on, deployed, and deliver real value to operators quickly. USG has expressed interest in making a substantial, broad investment in diverse industry and academic partners, building on OCUDU for 5G/6G, special capabilities, and Future-G initiatives. Our accelerated compute and AI-RAN functionality will help position OCUDU as a critical resource to accelerate defense and commercial innovation, improve speed and performance, deployment, not only in today’s standards, but beyond them. By building AI-RAN on open source, we can all move faster in the pre-standardization and tailored capabilities space.
Speed as a Competitive Edge: Software Standards Vs. Committee Standards
While nations and companies continue to engage in complex many-party 3GPP discussions to define 6G and ensure global interoperability, OCUDU developers can run faster, already implementing and testing 6G concepts on an interoperable platform, and helping to test what future standards may be. This shift from purely standards-driven to open source-driven development is one of OCUDU’s greatest strategic advantages. We need innovation to move faster, at the speed of software.
As this common software platform emerges, standardization may shift from solely official standards to de facto software standards—moving faster and better emphasizing innovative U.S. capabilities and software speed. Both SRS and DeepSig’s advancements in OpenRAN stack maturity, accelerated compute enhancement, and AI-RAN reference designs specifically enable this velocity with faster coordination, iteration, validation, and adoption. New AI models will deploy without hardware or foundational stack changes, novel waveforms can be deployed and tested without RF modifications, ISAC algorithms evolve based on real-world feedback – and all of these can ride on the evolution of open software, reference designs, and interfaces.
OCUDU acts as both an innovation and testing ground as well as a mature and performant deployment platform, giving 3GPP groups the ability to validate current and future standards and features in live network conditions, and to coordinate designs prior locking down standard functions. AI-RAN Alliance members test and validate blueprints for new use cases before specification. But more critically, OCUDU enables capabilities beyond committee imagination—innovations that become future standards through proven deployment and software consensus rather than solely isolated committee consensus. This model allows for the U.S. and others to lead through innovation and implementation, rather than negotiation alone.
Robust Deployment: From Lab to Production
OCUDU bridges the gap from experimentation to production deployment. The same software that runs on a researcher’s laptop for algorithm development scales effortlessly to carrier-grade deployments, handling millions of connections. This consistency across environments—from ARM-based edge devices to large x86 or GPU datacenter setups—removes the traditional divide between prototype and production.
The platform’s strength comes from its layered architecture and hardware flexibility. Core CU/DU functions can run on reliable x86 servers for proven stability, while heavy or AI workloads leverage GPU acceleration. Both target virtualized environments where resources can be rapidly shared, scaled, and adjusted in response to failures or between diverse workloads and applications. This software virtualization and decoupling from specific hardware, combined with orchestration and automation, enables zero-downtime updates, hardware maintenance, efficient scaling and load balancing, and value-driven compute hardware evolution.
Real-world deployments demonstrate this flexibility. A smart port deploys OCUDU on ruggedized ARM servers for environmental resilience, connecting hundreds of IoT sensors and autonomous vehicles. A financial trading firm runs the same software on high-frequency Intel processors for ultra-low latency connectivity to trading systems. A rural connectivity provider leverages cost-effective compute hardware to bring affordable 5G to underserved communities, and a large-scale mobile network operator leverages GPU compute to efficiently handle large Massive MIMO and AI-RAN workloads for spectral efficiency and high energy efficiency. Each deployment optimizes for different constraints—power, performance, or price—while running the same proven OCUDU software stack, and allowing for diverse software features, vendors, and specializations.
Unleashing New Use Cases Across Industries
OCUDU’s flexibility allows for broad growth of use cases that build upon the foundation of 5G, and were once impossible or too costly for many to implement or deploy. By offering open, customizable CU/DU implementations and acceleration, innovative applications across numerous industries, verticals, and government applications can be built and deployed with less barriers to entry and adoption.
In manufacturing, industrial use cases and logistics, OCUDU enables fully automated facilities where robots communicate over reliable 5G, while AI monitors production quality in real-time leveraging localization, positioning, and smart scheduling. The same network can also handle computer vision, large language models, and agentic workloads helping automate complex tasks, run predictive maintenance algorithms, and maximize our ability to reliably deploy autonomy—all on shared CPU and GPU infrastructure. Smart cities use OCUDU to build urban nervous systems: traffic lights that adjust to flow, emergency services that navigate through AI-optimized routes, and environmental sensors that trigger automated responses to pollution events.
Numerous industries including autonomous vehicles, internet of things and machines, and healthcare all see revolutionary possibilities. Autonomous vehicles may provide transport for humans, machines, delivery, and security while diverse IoT devices may share machine-type communications broadly or from low power reduced capability devices, improving the efficiency, autonomy, convenience, and security of modern life. Hospitals and critical infrastructure facilities can use OCUDU-based networks across critical robots to AI-powered monitoring, prioritizing latency-sensitive, and critical communication, opening the door to Diverse and carefully managed quality of service and integration of AI with the network.
The platform also enables the development of entirely new business models. Network operators can more easily virtualize and share workloads on their computing infrastructure and offer “AI-as-a-Service” capabilities at the edge alongside connectivity, monetizing excess GPU or compute capacity. Enterprises can deploy hybrid networks that provide both private 5G and edge computing from a single infrastructure investment. Researchers can prototype 6G concepts like AI-Native communications, learned beamforming communications systems, ISAC, spectrum sharing and digital twins while maintaining backward compatibility and load sharing with existing devices. These use cases demonstrate that open CU/DU implementations don’t just replicate existing capabilities—they unlock entirely new possibilities—and provide a powerful foundation for acceleration.
Building the Future: Open Source Speed Delivers Mission Goals
OCUDU embodies more than just technological innovation—it’s a new approach for maintaining technological superiority. By providing open CU/DU implementations, we unlock AI-RAN capabilities that make current standards broadly accessible and surpass current standards. This openness enables rapid innovation, experimentation, quick software integration, AI deployment, and 6G evolution at unprecedented speed.
The USG recognizes the opportunities ahead and is tentatively planning initiatives to drive innovation and capabilities in 5G/6G, ISAC, AI-RAN, dynamic spectrum, and tactical communications. OCUDU is emerging as the primary platform for advancing next-generation capabilities and rapidly deploying them in real-world validation. DeepSig’s role in this effort is to help accelerate these innovations onto OCUDU, ensuring that concepts relying on rapid innovation and a vibrant open-source ecosystem alongside a mature RAN, move new use cases much more rapidly into implementation and impact both within commercial markets and the hands of soldiers.
OCUDU’s permissive BSD3-based license unlocks broad commercialization without encumbrance, fueling opportunities across countless verticals and use cases. Together with the Linux Foundation and the global developer community, we are reshaping the rules of the international NextG innovation. The future of wireless will not be decided in committee rooms but written and proven in code, and deployed leveraging enabling open source. Whether enabling today’s networks or driving tomorrow’s breakthroughs, OCUDU is the platform where software speed and customization outpace standards speed.
