Exclusive First Release: Noematrix, Co-founded by Lu Ce-Wu, Completes Angel and Pre-A Rounds of Financing
In recent developments, Noematrix, a startup in the field of embodied intelligence, has announced the successful completion of its Angel and Pre-A rounds of financing, raising a cumulative amount in the hundreds of millions. The company, established in November 2023, is strategically incubated by Non夕 Technology Group.
The latest round of financing was co-led by Prosperity7 Ventures and GF Xinde, with participation from ZeYu Capital, Innovation Works, Qiyi Chuangtan, PlugandPlay China, and MFund Magic Capital. The Angel round was led by Xiaomiao Langcheng, with MFund Magic Capital, Jinghai Zhan Yue, and Infinite Fund SEE Fund also contributing.
The funds raised will be directed towards product development, commercial expansion, and talent acquisition.
(Source: Noematrix)
The primary challenge in embodied intelligence is the effective representation of the physical world and the interactions between robots. Existing large models for natural language and visual language struggle to fully capture the complexity of the real world, with reliance on visual and language modalities proving insufficient for comprehensive descriptions. Additionally, the data required for training embodied intelligence models is costly to acquire, inefficient, and lacks a unified standard.
Noematrix has addressed these challenges by developing a unique technical approach and strategic direction, unveiling an embodied intelligence brain system centered around force. By rapidly iterating on general skills, the company has achieved an efficient and cost-effective solution.
Most embodied intelligence architectures rely on visual language models for learning representations and reasoning about the environment and tasks, using imitation or reinforcement learning to optimize decisions based on trajectory positions. Noematrix, however, leverages its expertise in force control to integrate multi-dimensional force modalities and control strategies into the learning space, creating a framework for embodied intelligence models based on physical common sense and behavioral decision-making.
The physical world model, through operation-oriented topological structure understanding, task-centered physical common sense understanding, and pose-shape joint training, offers high accuracy and low training costs while addressing the challenge of manipulating deformable objects. The robot behavior model, centered on force and objects, further reduces the training space, achieving comprehensive force-position hybrid decision control with complete training information, high transferability, high precision, and robustness.
The joint training of these two-tiered architectures significantly reduces data requirements and enhances practical applicability.
Noematrix’s technological advancements are closely tied to the founders’ deep academic and industry experience in AI and robotics. Co-founder Lu Ce-Wu, currently a professor at the Department of Computer Science and Engineering at Shanghai Jiao Tong University, is one of the earliest scholars in China to research embodied intelligence. During his postdoctoral research at Stanford University’s AI Lab, he was mentored by Prof. Fei-Fei Li and Prof. Leo Guibas, an IEEE Fellow.
Lu Ce-Wu was named one of the 35 Innovators Under 35 by MIT Technology Review in 2018 and received the 2023 Science Exploration Award for his contributions to embodied intelligence. His team pioneered the AnyGrasp grasping algorithm, achieving human-level performance for the first time, and developed the HAKE human behavior knowledge engine. They have published over 100 papers in high-level journals and conferences, including Nature and Nature Machine Intelligence, and Lu Ce-Wu is the only Chinese scholar invited to join the Open X Embodiment Dataset international joint embodied intelligence large-scale dataset project.
Co-founder Wang Shi-Quan holds a Ph.D. from Stanford University and is a visiting professor at Shanghai Jiao Tong University and a member of the Shanghai Science and Technology Association. He is dedicated to the development of general intelligent robots and founded Non夕 Technology in 2016, pioneering a humanoid general robot technology path based on ultra-precise force control and hierarchical intelligence.
During the World Artificial Intelligence Conference in July 2024, Noematrix demonstrated its embodied brain system in a variety of tasks, including floor cleaning, desk tidying, manipulation of deformable objects like folding clothes, and food preparation such as peeling cucumbers.
(Source: Noematrix)
Noematrix officially launched its self-developed Noematrix Brain, a system centered on versatility and intelligence, providing a comprehensive, integrated hardware and software platform and toolkit covering the entire lifecycle of intelligent agent applications, from development and testing to deployment and validation.
By optimizing hardware selection and accelerating algorithms and software architectures, Noematrix offers a hardware-software integrated brain system to ensure outstanding performance and compatibility with various embodied intelligence devices. Standardized solutions, combining hardware and software expertise, are introduced as an application development platform, facilitating efficient and convenient embodied intelligence application development for developers.
(Source: Noematrix)
Through a unified model architecture, in-depth business judgment, and scene screening, Noematrix has identified that breakthroughs in common skill-related areas are crucial to addressing market pain points. The company has introduced the AnySkill series of universal skill products, adaptable to various embodied intelligence platforms. These products, all tested on actual machines, demonstrate high environmental adaptability, strong object generalization capabilities, and significant market application value.
The algorithmic capabilities of the Noematrix embodied brain will be significantly enhanced through the extensive accumulation and combination of universal skills, expanding its range of application scenarios. Noematrix has preliminarily validated the progressive path from technology development to commercialization with the AnySkill series, forming a sustainable positive cycle.
This validation not only demonstrates the practical effectiveness of the technology but also highlights its potential for continuous development. As technology advances and commercial applications expand, Noematrix’s embodied brain is poised to make a greater impact in various fields, driving the widespread adoption of embodied intelligence technology.
Views: 2