Research & Innovation
Looking to the future, Huawei’s innovation will transition as we move from an Innovation 1.0 era to an Innovation 2.0 era. In Innovation 1.0, we have focused on innovations in technology, engineering, products, and solutions to address customer needs. In Innovation 2.0, driven by our vision, we will expand our efforts to seek breakthroughs in basic theories and develop new basic technologies.
Open Innovation: Aggregating Wisdom and Innovation of Humanity to Light the Way Forward for the World and Huawei
In Innovation 1.0, we have focused on innovations in technology, engineering, products, and solutions to address customer needs and challenges. This is about going from 1 to N. The primary goal is to help our customers and partners become more competitive, increase their revenue or reduce their costs, and enable them to achieve greater business success. In the past, Huawei made many engineering and technological innovations in wireless, optical networks, and smart devices. These innovations generated a large amount of commercial value for our customers and tremendous social value.
In Innovation 2.0, based on our vision for the intelligent world we are entering, we will aim to break the bottlenecks in theories and basic technologies that have hindered the development of ICT. In this stage, we will focus on theoretical breakthroughs and inventions, which means going from 0 to 1.
Philosophy: Open innovation and inclusive development
Theoretical breakthroughs and inventions involve many uncertainties, meaning innovation should not be a closed- door activity. The philosophy of Huawei’s Innovation 2.0 will bring together universities and research institutes and connect the academia with businesses, to allow all types of global scientific research resources and talent to come together in the pursuit of innovation.
Methodology: Vision coupled with technological breakthroughs
Driven by our shared vision for the future, we will delve deep into the challenges the future intelligent world may bring regarding the way we live, work, play, and stay healthy. With these challenges in mind, we will try to determine the business opportunities and technological development directions which we should pursue, and endeavor to incubate new industries. On this basis, we will make bold assumptions and plan our future technological elements.
Focus: Exploring disruptive technologies throughout the information lifecycle
We have been exploring future-oriented technologies that can be applied to the entire process of information generation, storage, computing, transmission, presentation, and consumption. In addition, we have ramped up efforts on research into disruptive technologies.
Strategic initiative: Collaboration with universities and technology investment
Collaboration between businesses and universities can accelerate the commercial use of research results. And with insights into the challenges, real-world scenarios, and needs shared by businesses, scientists can also better decide what directions their future research should go in. To better implement Innovation 2.0, Huawei will adopt a number of approaches, including supporting the research programs of universities and research institutes, building its own labs, and investing along multiple technological paths.
Innovation has been the very foundation for Huawei’s survival and growth over the past 30 years. We will continue redoubling our efforts in exploring future-oriented cutting-edge technologies and basic research, with an annual investment of US$3 billion to US$5 billion. Today, Huawei has approximately 15,000 employees engaged in basic research, including over 700 PhDs specializing in mathematics, over 200 PhDs specializing in physics and chemistry, and over 5,000 PhDs specializing in engineering.
Meanwhile, we have established research and innovation partnerships with over 300 universities and more than 900 research institutes and firms. Following the principle of open innovation, we will leverage global innovation resources and attract professionals from around the world to collaborate on research. We will consider the challenges facing businesses and academic insights, as well as the confidence of venture capitalists to jointly innovate. The concerted push for innovation will allow our innovation accomplishments to be shared across both our industry and society as a whole, lighting the way forward for the world and Huawei.
Key Innovations Enable a Fully Connected, Intelligent World
Huawei is continuing to focus on research and innovation to seize the opportunities and address the challenges of the future intelligent world. This will widen the “de Laval nozzle”, the rocket engine that will drive us forward. We are also bringing top minds from around the world to join us, while openly innovating and incubating new theories and key ICT technologies to share with the industry. Together we can jointly serve the information society.
Redoubling efforts in basic research to make theoretical breakthroughs, fueling the development of the intelligent world
The rapid development of communications technology over the years continues to drive communications channel capacities closer to the Shannon limit. The whole industry is continuing to search for breakthroughs in underlying technologies, so that communications technologies can keep pace with the development of the intelligent world. In 2019, we continued our deep exploration of communications theories, aspiring to redefine a wireless network architecture for the time when the Shannon limit is broken. Specifically, we:
- Were the first in the industry to define the 10 grand challenges facing basic research and the challenges we foresee as we stand at the threshold of the intelligent world. These challenges include problems related to privacy, information models, channel capacity, and approximate computing and optimization. We have also been working with top scientists from around the world to tackle these challenges.
- Pioneered a theoretical model for trustworthy networks in the age of AI, a dynamic network data traffic forecast model, and a theoretical model for distributed dynamic sliced networks. These models have significantly improved network robustness, forecastability, and utilization, and driven active control efficiency up by 56%.
- Achieved new breakthroughs in key theories and scheduling algorithms for next-generation massive Multiple Input Multiple Output (MIMO) that are applicable to future wireless networks, increasing edge user experienced data rates by at least 30%.
- AI has become a major driver of revolutionary changes to computer architecture. In this area, we:
- Attached importance to basic research, including both theoretical breakthroughs and technological innovation.
- Published more than 80 papers on AI, one of which was the only paper to win the ACL 2019 Best Long Paper award.
- Were the first to propose AdderNets in the deep learning field, which only contains additions and breaks the traditional framework of massive multiplications. This new method features software-hardware co-design and thus can achieve high performance on computer vision tasks and potentially save over 10 times the current requirements in terms of energy consumption and chip area. We have also made the testing code of AdderNets public to the community, aiming to explore the next generation of AI accelerators together with academia.
- Created the industry’s first framework of causal structure learning in replacement of the conventional local search method. This novel framework is based on reinforcement learning and can achieve state-of- the-art accuracy on academia benchmarks. Our paper on this framework was accepted to ICLR 2020 and ranked first (parallel with other submissions).
- Pioneered the proposal of comprehensive explainable AI measures based on computational metrics, and designed novel explainable methods from data-driven and knowledge-driven aspects, in order to construct complete and explainable AI benchmarks in the industry.
- Supported the implementation of hardware design space exploration technology with Pareto optimizations, the first of its kind to employ the Design Space Explorer (DSE) which boosts chip design efficiency and performance.
Innovating and inventing to drive the industry forward
Advanced Wireless Networks
Today, we have entered the 5G era and embarked on 6G research. Huawei is exploring new directions for 6G. To this end, we:
- Began research into the basic theories of 6G, including research into new air interface technologies, new network architectures, and key enabling technologies.
- Initiated extensive open collaboration projects with other industry players and pushed the industry to build a consensus on 6G.
To maintain our leadership in this area, we:
- Continued efforts to push the limits of fiber capacity and created industry-leading super-core transmission sites, which will ensure that fiber capacity can double every 36 months, even in the ultra-broadband age.
- Addressed the difficulties surrounding key technologies like high-performance optical algorithms for long-haul transmission at data rates of 400 Gbit/s per wavelength, high-speed optoelectronic components, and L-band wide-spectrum optical amplifiers, extending the transmission distance to more than 1,600 km and increasing the capacity of a single fiber to 48 Tbit/s.
- Proposed an optical cross-connect (OXC) architecture that supports wavelength-based flexible scalability to a super large capacity of 10 Pbit/s.
In this area, we:
- Reduced repeated work order assignment by using intelligent identification and more accurate fault location to ensure only a single work order is placed for each fault.
- Performed forecasts using spatial-temporal algorithms and dynamic equipment adjustments on wireless networks, resulting in a 15% improvement in energy conservation during pilot programs and reducing the cooling power consumption for data centers with a higher power usage effectiveness (PUE) by between 9% and 17%.
- Worked with other industry players to release a white paper entitled Autonomous Networks: Empowering Digital Transformation For The Telecoms Industry on the five levels of Autonomous Driving Network (ADN), as previously defined from L0 to L5, and their technical implementation paths.
- Continued research into ADN technology, to make intelligent, automated, self-optimizing, self-healing, and autonomous network operations and maintenance (O&M) a reality.
In this area, we continued to build on our full-stack capabilities ranging from computing power, operators, and the framework to application enablement, lowering AI application threshold and providing affordable AI capabilities for industries. Specifically, we launched:
- The Ascend 910, a proprietary AI training processor packed with a Da Vinci 3D cube architecture; and
- MindSpore, an all-scenario AI computing framework across devices, edge, and cloud.
In this area, we:
- Used AI to redefine the smartphone photography experience by being the first in the industry to deliver a staggering ISO 409,600, which enables crystal-clear night-time images, as well as 10x hybrid zoom and 50x digital zoom.
- Unveiled the industry’s first-ever high-precision segmentation algorithm, enabling high-precision foreground extraction (even for fine details like hair) and achieving the best bokeh and video effect in the industry.
- Introduced AI to sharpen our edge in device video, by improving scene identification and video
- enhancement, and delivering brand new experiences such as shooting ultra-slow motion, as well as ultra-wide-angle, ultra-low-light, and time-lapse videos.
Media CODEC Standards
To promote technological advancement within the ultra- high-definition video industry, we actively participated in setting standards in standards organizations such as the Moving Picture Experts Group (MPEG) with the ISO/IEC, ITU-T SG16, and China’s AVS by:
- Submitting more than 500 technical contributions to different standards organizations; and
- Acting as a major contributor to standards in fields such as Versatile Video Coding (VVC, or referred to as ITU-T H.266), MPEG-5 EVC, and AVS3.
In this area, we have always continued our efforts to pursue innovation and breakthroughs, so as to build our leadership in underlying software. In 2019, we:
- Unveiled the world’s first-ever AI-native and enterprise-class database with an AI kernel.
- Launched OpenEuler, an open source server operating system, based on a RISC and multi- core architecture. OpenEuler features three-level intelligent scheduling, which can reduce the multi- process concurrency latency by 60% and improve the performance of web servers by 137% through intelligent and automated planning, giving full play to the strengths of our Kunpeng and Ascend processors in computing power.
To increase the trustworthiness of our products and services, we have:
- Ramped up efforts to protect the digital sovereignty and privacy of customers, and lay a solid foundation for building trust in the digital society.
- Joined hands with industry players and academics from around the world to actively drive the development of trustworthiness theories, standards, and specifications.
- Become a major player and contributor of trustworthiness-related work groups in the ISO, Internet Engineering Task Force (IETF), 3rd Generation Partnership Project (3GPP), and European Telecommunications Standards Institute (ETSI).
- Actively sought certification for our trustworthiness from industry organizations, like gaining the CC EAL5+ certification for the HongMeng Kernel, making it the first commercial microkernel in the industry to receive this certification, and the CC EAL4+ certification for EulerOS.
- Released the Huawei Trustworthiness Framework.
- Continued our research and exploration into cutting- edge technologies, including AI trustworthiness, post-quantum security, differential privacy, and the digital trust mechanism.