The performance goals for 5G are ambitious, impressive, and will be world-changing. Achieving them while scaling network density, capacity and evolving to open and distributed architectures is a complex challenge. Numerous traditional algorithms and radio network design, deployment and management approaches are showing their limits.
New applications, industrial and IoT applications, and ever-increasing customer demand for mobile data is driving the urgency to move from 4G to 5G to increase capacity and lower latency. While 5G promises significantly enhanced and pervasive user experiences and device connectivity. Many core wireless signal processing approaches are over 20 years old and still used today. Moore’s Law has continued to reduce the rack-scale systems of the 1990s to our present-day smart phones and cloud infrastructure, but the core algorithms have not evolved at the same pace. The result is 5G systems often consume dramatically more power than desired, achieve lower data rates than planned, and leave remaining network density and performance untapped. DeepSig addresses these problems by integrating machine learning and AI throughout core components within the radio access network to dramatically reduce power consumption and improve performance. This is a fundamentally different and far more performant data-driven approach than many of the existing AI-based solutions which solely focus on upper layer resource allocation layers. DeepSig has productized mature drop-in software solutions in our OmniPHY™ solutions and is working with 5G infrastructure and industry partners to provide unprecedented levels of performance enhancement in current and next generation RAN systems.
Our OmniPHY software leverages key advances made in machine learning over the past few years which have transformed fields of computer vision, autonomous vehicles, natural language processing, and voice recognition but fundamentally applies these concepts within key baseband processing tasks which suffer from high complexity. We leverage the explosion of open software and ideas that have helped transform these industries from companies such as Facebook and Google to accelerate and enable our software solutions prototyping, implementation and development. Our core driving philosophy is that today insufficient wireless models and algorithms are used throughput the wireless stack, and that more data-driven machine learning approaches can adapt better to real world effects enhance system performance dynamically and leading to improved capacity, density, resilience, and efficiency. Instead of relying on engineering approximations and simplifications in wireless systems, our deep neural network-based signal processing solutions embrace complexity and data and learn near optimal and low complexity solutions for to improve key performance metrics such as data throughput, multi-user capacity, latency minimization and power consumption reduction. This approach can increase the overall quality of service, reduce CapEX and OpEx, and make improve the performance of the radio access network across a wide range of components and effects.
5G today employs industry standard air interface waveforms, such as the “5G New Radio Air Interface (5G-NR)”, which are developed by the 3GPP International Committee. Our OmniPHY machine learning signal processing solutions are split into two classes: (1) Transparent 3GPP compliant RAN system enhancements which fully function with today’s standard releases with zero changes (2) Beyond-5G radio technologies where OmniPHY can adapt the fundamental waveform beyond this specification to drive further performance enhancements and improvements (candidate technologies for future standard use). Thus, to achieve performance improvements in 5G, OmniPHY substitutes AI and novel software signal processing blocks within existing RAN components while maintaining full compatibility with the 3GPP 5G standard and Open RAN architecture standards.
This process, replacing conventional analytic algorithms that suffer from with model or complexity drawbacks with AI-based equivalents that use machine learning to adapt and improve after deployment brings significant computational and power benefits several key areas of the 5G stack, allowing for operating cost reduction and relaxation of hardware fidelity requirements, to reduce capital cost by increasing cell and user capacity on current hardware and spectrum resources.
DeepSig’s enhanced algorithms within the 5G RAN help reduce power, reduce component cost, increase device density and performance, and automate tuning, operation, and deployment of diverse open 5G RAN hardware in terms of both OpEx and CapEx.
DeepSig is investing heavily in the application of AI for 5G and other wireless technologies. We are rapidly developing and validating these capabilities tightly coupled with industry standard NR RAN software implementations. Real-time demonstrations of our solutions running over the air on NR test equipment within commercially deployable O-RAN stacks are available in wideband NR configurations. If you would like to learn more, please get in touch using the contact form below to understand how we can make your 5G NR systems perform faster, more efficient, secure and autonomous.
Our core capability interoperates with leading O-RAN Upper L1 PHY Software capabilities already deployed in production in order to better exploit channel information and reduce processing requirements. These together serve to improve multi-user capacity and experience while scaling to support more sectors on existing edge processing hardware solutions within the Distributed Unit.
By learning to enhance spatial processing and combining in the field, with data-driven techniques, we reduce computational and power requirements for scaling 5G RAN systems to many-antennas and many-user environments and further improve performance under non-ideal conditions in complex urban environments where propagation, interference, distortion and mobility can all be very hard to model and pre-optimize for in closed form.
We have demonstrated machine learning and data-driven approaches to enhance Radio Unit (RU) performance by better compensating for non-ideal hardware effects, interference, and non-linearity. By efficiently transporting and adapting the 5G-NR signal within commercial RU solutions, we can reduce component and operation cost to help scale ORAN radio systems.
.png "image (2)")