Recently, the results of the 2024 “Invention and Entrepreneurship Award” Innovation Award by the China Invention Association were announced. AICT, in collaboration with Beijing University of Science and Technology, the Institute of Automation of the Chinese Academy of Sciences, and other entities, won the first prize for their jointly submitted project “Software-Hardware Collaborative Edge AI SoC Chip Technology and Application.” This award is a significant achievement in AICT’s ongoing efforts to promote innovation and the application of AI technology in the industry, marking another milestone in high-accuracy AI (HAI) research and development.
The “Invention and Entrepreneurship Award” aims to encourage invention and innovation, promote the transformation and implementation of invention achievements and patented technologies. Approved by the Ministry of Science and Technology and registered with the Office of National Science and Technology Awards, it consists of four sub-awards: Project Award, Person Award, Achievement Award, and Innovation Award, holding widespread social influence and authority in the field of invention and innovation nationwide.
Dr. Xiang Yanping, who is CEO and Chairman of the Technology Committee of AICT, expressed that this award is a recognition of the company’s R&D team’s relentless dedication to innovation. In over nine years since its founding, AICT has remained focused on developing and applying HAI products, and its HAI technology and solutions have been widely applied in areas such as perception robots, intelligent traffic lights, smart roads and autonomous driving, vehicle-road-cloud integration, low-altitude air transportation, and space intelligence. AICT will continue to uphold its mission, adopt a more open mindset, and take firm steps toward exploring deep integration of AI with various industries, bringing more innovative results to society and contributing to the development of smart cities in China and worldwide.
The development of AI and integrated circuit technology is of strategic importance to national security and development. Edge AI system-on-chip (SoC) technology, the core component of intelligent edge devices, faces challenges such as limited computing resources, energy supply constraints, and complex design processes, along with low accuracy of intelligent algorithms, high chip power consumption, and limited self-designed tools. Software-hardware collaborative design is the key technology to address these challenges and is essential for the research, production, and application of edge AI SoC chips.
To tackle this “bottleneck” issue, AICT, together with Beijing University of Science and Technology, the Institute of Automation of the Chinese Academy of Sciences, and other entities, established a joint project team for collaborative innovation. This project addresses major needs in AI applications, inventing critical software-hardware collaborative technologies for AI SoC chips, developing efficient edge AI SoC chips, and achieving large-scale applications, which significantly promote the development and application of edge AI technology.
During the project’s development, AICT and its team members jointly invented a hardware-adaptive joint optimization method for deep neural network model pruning and quantization, and proposed a high-accuracy lightweight intelligent perception algorithm adaptive to chip hardware, effectively solving the issues of limited computing resources and low accuracy of lightweight intelligent algorithms. The team also invented a chip memory architecture optimized for intensive computation and a software-adaptive design method for neural network acceleration units, building an efficient AI SoC chip design system based on software-hardware collaboration, effectively addressing issues of limited energy supply and high chip power consumption. Additionally, the team invented a rapid layout and routing method based on knowledge and reinforcement learning, developed a cross-layer collaborative high-efficiency chip design tool for EDA layout and routing, effectively solving the complex design process and lack of self-designed tools for AI chips.
The “Software-Hardware Collaborative Edge AI SoC Chip Technology and Application” project is characterized by high technical complexity, significant development challenges, and strong innovation. It possesses independent intellectual property rights, with core technology that is autonomously controllable and reaches an internationally advanced level. The hardware-adaptive joint optimization method for deep neural network pruning and quantization, along with the software-hardware collaborative efficient AI SoC chip design approach, are at the forefront globally.
Until today, the project’s achievements have garnered multiple national invention patents and numerous high-quality academic publications. The project has developed over ten edge AI SoC chips, with cumulative production and sales in the tens of millions. The technology has been widely applied on a large scale in fields such as intelligent transportation, smart security, intelligent hardware, and smart grids, yielding significant economic and social benefits. Especially in response to the needs of intelligent transportation, the project developed a series of AI devices and established city-scale intelligent road network collaborative perception and management systems, which have been implemented in over 50 large and medium-sized cities including Beijing, Zhangjiakou, and Guangzhou.
Media Contact
Company Name: Artificial Intelligent Interconnection Technology Co., Ltd.
Contact Person: Wu Jinna
Email: Send Email
Country: China
Website: http://www.zhihuihutong.com/