Research & Innovation
Digitalization and carbon neutrality are two of the world's most important current topics. They are both making a deep and lasting impact on the ICT industry. The global digital economy is growing rapidly, and the demand for digital products and services has outstripped expectations. However, supply is struggling to catch up, because both Shannon's theorem and the von Neumann architecture have run into severe bottlenecks. In 2021, Huawei invested even more in research and innovation, as part of our greater efforts to sustain development in the future. We are sparing no efforts as we explore the endless frontiers of science and technology. We are also devoting efforts to identifying the needs of different industries and overcoming global challenges. Guided by our vision of a fully connected, intelligent world, we will work openly with the global scientific community to explore new theories, architectures, and technologies, which will support the ongoing development of the industry.
In 2021, Huawei ramped up investment in basic research into fundamental theories about communications, computing, AI, and many other fields.
We continued to make advances in the fundamental theories on wireless communications, pushing our algorithms ever closer to the theoretical Shannon limit. Our innovations in the theory and algorithms of sparse superposition codes have enabled us to derive the Shannon limit for a fast-changing coding matrix, helping us to substantially simplify the design of transceivers.
We published our 6G vision, defining new services, use cases, and technological trends for the future, as part of the industry-wide efforts to create a consensus around 6G.
We created a new, trustworthy architecture that combines full-spectrum wireless communications and sensing with native AI, incorporating the New Radio (NR) framework, ultra-dense satellite networks, and terrestrial cellular networks. This new architecture will support intelligent 6G connectivity.
We continued to research new architectures and new components for wireless transmission, reshaping conventional cellular network designs, and proposed a capacity-centric architecture for future wireless networks. We were able to calculate the theoretical maximum network capacity and defined the capacity changes to help increase the coverage and capacity of future wireless networks combining both high- and low-frequency resources, setting the wireless network industry on a stable footing for the future.
We advanced optical communications with new approaches to non-linear waveform designs, making important progress on the industry-wide problem of non-linearity in fiber and significantly extending transmission distances.
We created a precision network traffic prediction model and refined the actuarial theory for network service level agreements (SLAs), enabling traffic models to make a leap in levels of precision, from the millisecond level to the microsecond level. This has made network performance much more predictable and boosted resource utilization.
We continued our research into storage coding technologies. We were the first to develop divide-and-conquer mapping and vertical horizontal recursion coding, which allowed us to achieve breakthroughs in low-complexity, incremental error correction coding, and increase the throughput of distributed storage under typical load distributions.
We have made new progress in AI image encoding and decoding technologies, and supported the launch of the JPEG AI project (ISO/IEC JTC 1/SC29) to develop the next generation of image standards.
Our AI-powered mixed audio codec architecture has been approved by the China Ultra-HD Video Industry Alliance (CUVA) and Audio Video Coding Standard (AVS), China's working group on audio and video codec standards, and will provide the technical foundation for their spatial audio codec standards.
Our dynamic HDR core technologies are helping the HDR Vivid industry standard achieve widespread acceptance in the industry. Over 10,000 hours of video content has now been delivered using HDR Vivid.
Left picture: Laurent Lafforgue is a world-renowned mathematician. Born in 1966, he won his first silver medal at an International Mathematical Olympiad (IMO) when he was just 18. At the age of 35, he received the Fields Medal, the mathematics equivalent of a Nobel Prize, for his outstanding contributions to number theory and algebraic geometry. In 2021, Lafforgue joined Huawei's Paris Research Center. The topos theory he works on is a highly abstract mathematical proposition that may point the way to new worlds of communications, computing, and AI. The company also set up a Lagrange Mathematics and Computing Research Center in Paris, France in 2020. This center aims to attract top-tier academics to be part of Huawei’s research programs and to nurture young researchers.
Right pictures: The Lagrange Mathematics and Computing Research Center
Focus, Persevere, Break Through
Huawei has spent decades investing heavily in R&D. In 2021, we once again witnessed how our "Focus, Persevere, Break Through" strategy drove innovation and rapid advances across the industry.
In this area:
We made breakthroughs in optical network technologies, such as new coding, signal processing, and optoelectronic components, supporting our evolution towards higher bit rate per wavelength over long haul.
We have a good grasp on key materials and manufacturing techniques for optical amplification, and verified the feasibility of ultra-wide spectrum transmission technology. We have also recast the optical backplane architecture and developed new technologies to address engineering challenges like high-precision, high-density optical connections in the C + L bands. These efforts have doubled transmission capacity per fiber.
We generated groundbreaking results in key technologies like high-power light source pooling and silicon photonics. This supports N x 100G, low-cost, high-density optical interconnection within data centers and maximizes the computing performance of their networks.
We made innovations in optical fiber sensing, spectrum detection, and micro-electro-mechanical system (MEMS) sensing algorithms and architectures, and promoted the adoption of optical networks to help vertical industries, including oil and gas, coal mining, and ports, go digital.
Carrier and Enterprise Networks
We made innovations in Ethernet technology. Specifically, this year we defined a simplified, low-latency data center network technology stack. When converged with Unified Bus (UB), a computing-native interconnections technology, this will make a UB-over-Ethernet (UBoE) solution possible, creating large-scale computing networks, each with more than 300,000 nodes, a switching element capacity of over 100 Tbit/s, and a static latency of less than 130 nanoseconds.
The interconnection of heterogeneous networks (ManyNets) is an ongoing and irreversible trend. In 2021, we continued our own research into new network protocols and a distributed forwarding architecture. We also expanded our collaboration with carriers, industry customers, and partners, and carried out cutting-edge pilot projects in new industrial Internet, IoT network access, smart cities, smart campuses, and more. All of these efforts are aimed at providing a deterministic connectivity experience to our customers, underpinned by intrinsic security and ultra-low power consumption.
Autonomous Driving Networks (ADNs) aim to achieve automation, self-healing, self-optimization, and autonomy, and their four most prominent features are advanced intelligent sensing, digital mapping, self-learning, and adaptive decision making.
Supported by our new algorithms, wireless networks can learn and predict variations in network load, and orchestrate the use of multiple energy sources to maximize the use of green energy, while also reducing electricity costs by at least 10% compared with human-controlled networks.
We made breakthroughs in network knowledge graph technologies, accumulated 1.1 million pieces of network O&M knowledge across 15 products, and created AI inference tools that provide solutions in seconds.
As part of our full-stack AI portfolio, we launched the dense foundation model Pangu-Alpha, which was the industry's first to be trained with 200 billion parameters. We also invented an adder neural network which can replace convolution operations for deep learning, significantly reducing required floating-point multiplication and increasing energy efficiency by 30% to 50%.
We took significant steps forward in parallel model training, deployment and scheduling algorithms, and communications technologies for ultra-large-scale deployment of AI models, honing our competitive edge in power consumption, performance, and cost.
Our groundbreaking work in optimized compilation, automated deployment, and efficient execution in heterogeneous distributed compute clusters has helped halve the cost of computing for AI and search.
We made huge leaps in high-bandwidth algorithms for medium detection, error correction codes for high-performance memory, and in the theory and algorithms for low-redundancy erasure coding. These advances helped boost the reliability of our computing and storage functions.
With sustained innovation in computational photography, computational optics, and the True-Chroma Image Engine, Huawei remains a leader in digital photography. We are continuing to work nonstop to deliver a better image and video experience for our users.
Our groundbreaking work in core technologies such as spatial computing, spatial video, and 3D reconstruction is helping create immersive augmented reality experiences.
The adaptive noise reduction technology in our open-fit earphones takes noise cancellation to new heights.
Our ultra-HD, low-latency video processing technology supports easy multi-screen collaboration, delivering a better viewing experience for users.
We were the first to use a phone form factor that features a wedged shape and multi-dimensional hinge design. This allows the two screens to fold in snugly against each other, preventing dust or grit from getting in and scratching the screens or obstructing the hinge. These innovations have been applied in multiple Huawei foldable phone models and helped us lead in this sector.
We developed an eight-channel, high-performance, high-sensitivity photoplethysmography (PPG) module and an intelligent algorithm to minimize interference when a smart watch is monitoring a wearer's heart rate in sports mode. In more than 40 different sports scenarios, our heart rate monitor delivers the most accurate results in the industry.
We have continued to research and innovate in this area. Our innovations in software architecture and full-stack systems optimization have substantially improved the utilization of hardware resources and sharpened our competitive edge.
In terms of operating systems, we achieved breakthroughs in technologies like deterministic scheduling and flexible resource isolation. This helps us meet demands for low latency in embedded applications, and has doubled resource utilization in typical cloud scenarios.
Thanks to great progress in multi-core parallelism and automated vectorization technologies, our BiSheng compiler deeply synchronizes software and hardware, helping our proprietary chips lead the industry in diversified computing performance.
We've also made substantial headway in core technologies for GaussDB, such as fully parallel, multi-core processing and AI-assisted self-optimization. This has honed our edge in high performance, high availability, and autonomous O&M, helping us expand our presence in seven key industries.
We are committed to open source and openness in software ecosystems. Our multiple-pillar technology architecture helps us build technologies and ecosystems in key segments like finance, public services, and enterprise solutions. The openEuler and HarmonyOS ecosystems are developing rapidly and more than 220 million Huawei devices now run on HarmonyOS.
We provide technologies to support app porting, including connectivity, rendering, AI, maps, search, data management, Kit frameworks, and our AR Engine. More than 3,000 apps have now been ported to Huawei operating systems to date and over 600 million active devices are using Huawei Mobile Services (HMS).
We have made sustained investments in research into trustworthy theories, technologies, and engineering practices.
Huawei works with industry peers to develop theories, standards, and specifications for trustworthy technology, and is a major contributor to trustworthiness-related working groups in the International Organization for Standardization (ISO), Telecommunication Standardization Sector of the International Telecommunication Union (ITU-T), 3rd Generation Partnership Project (3GPP), European Telecommunications Standards Institute (ETSI), Internet Engineering Task Force (IETF), and other standards organizations.
We are dedicated to enhancing our own software engineering capabilities and the trustworthiness of open source software. To this end, we are researching trustworthy programming languages and developing new engineering capabilities. Huawei is also one of the five founding members of the Rust Foundation.
We research technologies for vulnerability management and open source to help more accurately and efficiently find open source vulnerabilities and ensure that code comes from trustworthy sources.
We are exploring cutting-edge technologies, including cryptography, AI trustworthiness, systems security, and privacy-enhancing computation, and applying the results of our research to our ongoing engineering practice to enhance product security and resilience.
We are studying engineering technologies like functional security and human factor engineering, bringing ICT technologies to all walks of life.