HomeArticle

Just one year after its establishment, the embodied intelligence enterprise "Qiongche Intelligence" has received hundreds of millions of yuan in financing led by Sequoia. | Early-stage project

胡依婷2024-12-05 16:48
The embodied brain obtains one hundred sets of orders for a single scenario.

Text|Hu Yiting

Editor|Yuan Silai

Folding clothes, peeling cucumbers, and even shaving, under the technological wave of embodied intelligence, the degree of fine operations that robots can achieve has gradually broken people's perceptions.

When multiple startups are competing on the same stage, the financing of embodied intelligence is also extremely hot. Recently, Sequoia China once again led the investment and invested in a one-year-old embodied intelligence startup, Noematrix.

Noematrix recently announced the completion of a Pre-A+ round of financing of several hundred million yuan, led by Sequoia China, with continuous investment from old shareholders Prosperity7 Ventures, Xiaomiao Langcheng, and Plug and Play China.

Noematrix is strategically incubated by Flexiv. It mainly researches and develops the basic model and system of embodied intelligence, and provides software and hardware tools and platforms covering the entire cycle of intelligent robot application development.

Since its establishment in 2023, Noematrix has completed three rounds of financing participated or followed by investors such as Xiaomiao Langcheng, MFund Magic Quantity Capital, Zeyu Capital, Chuangxin Gongchang, and Qiji Chuangtan.

One year after its establishment, it has released the Noematrix Brain and initially established a complete set of product systems including the entire process of data collection, model training, and deployment verification. Among them, the Embodied Brain has obtained hundreds of orders for a single scene, and in the end-to-end training of multiple complex meta-abilities, it can take into account the generalization and landing performance requirements, expanding the boundary of the application ability of the robot on the real machine.

In terms of technical routes, relying on the accumulation in force control technology, Noematrix introduces the multi-dimensional modality of force and control strategy into the learning space, constructing an embodied intelligence model framework system based on physical common sense and behavioral decision-making, driven by its two self-developed large models of embodied intelligence, namely the Physical World Large Model and the Robot Behavior Large Model.

Based on the two large models, Noematrix independently developed the "Force-Centered" Embodied Brain system, Noematrix Brain, which can achieve an efficient and low-cost solution by quickly iterating common skills such as grasping and placing.

In terms of products, in addition to providing a brain system that integrates software and hardware, Noematrix also launched the AnySkill universal skill series products that can be adapted to a variety of embodied intelligence ontology platforms. Currently, it can achieve four skills: grasping, folding of flexible objects, scraping, and opening.

AnySkill Embodied Intelligence Atomic Skill Library, Image Source: Enterprise

During the World Artificial Intelligence Conference in July this year, Noematrix announced the launch of the Noematrix Brain, and at the same time, its real machine completed the demonstration of multiple task processes in an open environment and in close proximity to the audience, including floor cleaning, desktop organizing, flexible object operations (such as folding clothes) and food material processing (such as peeling cucumbers), showing the universality and robustness of its products.

While achieving a variety of basic skills, Noematrix will be applied to practical scenarios such as retail operations, scientific research and education, home services, logistics picking, and medical drug sorting.

In addition, Noematrix also transforms the software and hardware experience into a standardized solution and launches an application development platform that integrates the ontology hardware, basic software framework, algorithm model, and supporting tools, which can meet various needs such as digital twin, data collection, model training, business development, and robot cluster management.

In terms of the founding team, the embodied intelligence brain product of Noematrix is led and developed by co-founder Lu Cewu, and the CEO of Flexiv, Wang Shiquan, also serves as a co-founder. Lu Cewu was once a postdoctoral fellow at the Stanford University Artificial Intelligence Laboratory, studying under renowned scholars Li Feifei and Leo Guibas, and is also the only winner of the Science Exploration Award in the field of embodied intelligence. Currently, the core management team of Noematrix has also recruited the former head of the humanoid robot of Pengxing Intelligence and business executives from Huawei to join.