news

tencent technology news (zhou xiaoyan) reported on september 6 that qiongche intelligence, a startup in the field of embodied intelligence, announced the completion of angel and pre-a rounds of financing totaling hundreds of millions of rmb. this round of financing was jointly led by prosperity7 ventures and gf securities, and participated by investment institutions such as zeyu capital, sinovation ventures, qiji chuangtan, plug and play china, and mfund capital.

at the world artificial intelligence conference on july 5, qiongche intelligence demonstrated a robot "cutting cucumbers" on site and also released a video of a robot "shaving" a human. as for the principles behind these two delicate tasks, qiongche intelligence founder lu cewu told tencent technology that this technology mainly involves two aspects: one is the robot's understanding of the task, and the other is the intelligence in control.

"if robots want to do things well, they need to understand 'force' feedback. for example, when humans twist a bottle cap, it involves the action of force. so when it comes to robots, it means making decisions on how to couple force and position. otherwise, if the force is too strong or too weak, they won't be able to shave cleanly."

embodied intelligence currently faces some challenges, such as how to better describe the physical world and how to interact with robots.

on the one hand, when it comes to describing the physical world, existing natural language and visual models are not ideal when dealing with complex problems in the real world, and it is difficult to fully understand the physical world relying solely on language and vision. on the other hand, a large amount of data is required to achieve human-computer interaction, and the data acquisition cost for training large embodied intelligence models is high, the efficiency is low, and there is no unified standard.

to address these issues, qiongche intelligence proposed to release a "force-centric" embodied brain system to provide low-cost solutions by iterating general skills.

common embodied intelligence systems mainly rely on visual and language models to learn and reason about the environment and tasks, and optimize robot decisions through imitation or reinforcement learning. qiongche intelligence introduces multi-dimensional force modalities and control strategies into the learning space, and builds an embodied intelligence framework that combines physical common sense and behavioral decision-making. this framework understands the physical common sense in operational tasks through a large model of the physical world.

in addition, qiongche intelligent launched a force-position hybrid robot behavior model, which combines "force perception" and position perception to make decision control more accurate. qiongche intelligent said, "this joint training method of the two-part model reduces data requirements while improving the efficiency and feasibility of practical applications."

during his postdoctoral period at the stanford artificial intelligence laboratory, lu cewu studied under professor fei-fei li and professor leo guibas (a member of the u.s. national academy of engineering). he is currently a professor at shanghai jiao tong university and is one of the earliest scholars in china to study the field of embodied intelligence.

lu cewu's team originally proposed the anygrasp grasping algorithm. in addition, the team's rh20t dataset is the first large-scale embodied dataset that includes force information. lu cewu was also invited to join the open x embodiment dataset international joint embodied intelligence large-scale dataset project.

wang shiquan, co-founder of qiongche intelligent, is a ph.d. from stanford university, a visiting professor at shanghai jiao tong university, and a member of the standing committee of the shanghai association for science and technology. he initiated the establishment of feixi technology in 2016. its adaptive robot products have been used in many industries such as industrial manufacturing, food processing, and medical services.

regarding this round of financing, lu cewu expressed that he would continue to promote the implementation of embodied intelligent technology and its industrialization.