Embodied Intelligence Technology Company DexForce Completes Strategic Financing Round

Technology Author: EqualOcean News, Jiahui Liao Editor: Jiaqi Li May 14, 2024 02:49 PM (GMT+8)

The financing round was led by Lenovo Capital.


Recently, DexForce, a company specializing in the development of highly versatile embodied intelligence technology, has completed a strategic financing round led by Lenovo Capital. The funds raised will be primarily used for product development, team expansion, and market expansion.

Founded in June 2021, DexForce is a national high-tech enterprise that focuses on Sim2Real technology. With long-term technical accumulation in 3D generative AI, multimodal large models, and 3D imaging, DexForce has created a product matrix that integrates both hardware and software based on Sim2Real, and has achieved commercial deployment in various scenarios. It is a leader in the large-scale commercial deployment of embodied intelligence.

Sim2Real is a technique that uses physical simulation of robotic operations and introduces various real-world disturbances related to the task (visual, physical, task description disturbances, environmental distractors, etc.). It then forms a massive amount of synthetic data with absolute precise standards through rendering, trajectory data recording, joint data recording, and other methods, using such synthetic data to train large models of embodied intelligence.

DexForce, with Sim2Real at its core, continues to build core technical capabilities at the foundation. It has created a unique DexVerse™ data and embodied intelligence simulation engine for data generation and large model training. This engine combines probabilistic modeling-based procedural generation simulation solutions with generative AI technology to address the shortcomings of existing technologies that cannot synthesize a variety of high-quality three-dimensional simulation data assets and the difficulty of human intervention, which leads to the generation not following real-world physical constraints. It achieves an efficient, zero-cost, and more realistic and reliable simulation data generation engine, with the ability to obtain a large amount of rich digital assets at a low cost, laying a solid foundation for the continuous low-cost generation of data. At the same time, combined with long-term accumulation in multimodal large models of embodied intelligence, DexForce has created an imaging perception suite based on the 3D VLA large model, which supports the phased realization of AnyGrasp/AnyManipulation in various industries, rather than custom task programming for different operational objects.

Shi Chenxing from Lenovo Capital said that DexForce has the most outstanding Sim2Real and large model-related technologies, which are the core base for achieving highly versatile embodied intelligence. Currently, DexForce relies on Sim2Real and related technologies, has a deep product accumulation in the field of embodied intelligence, has created a product matrix that integrates both hardware and software, including the Sim2Real AI engine based on 3D generative AI and the imaging perception suite based on the 3D VLA large model, and is in a leading position in the commercial deployment of embodied intelligence.

In addition, by empowering more mature robotic arms/robots, there have been good landing applications in industries such as semiconductors, automotive, and photovoltaics. The DexForce solution can effectively save a lot of deployment time, reduce costs, and improve stability and versatility compared to traditional technologies, supporting flexible production.

Jia Kui, the founder of DexForce, revealed that the company will follow the technical development trajectory of versatility in the future, successively covering scenarios from semi-structured to unstructured. Typical industrial/cooperative robotic arm scenarios will gradually achieve embodied intelligence imaging, perception, and control systems that support (close to) grasping/operating any objects and scenes, robot controllers that support rapid automatic task deployment on various robots, and autonomous robot systems capable of completing single and multi-round tasks independently. After the above is achieved, it will also open up the possibility of landing general-purpose/humanoid robots in a broader range of scenarios, including households.