news

Live coverage of 2024 WRC | Walking on complex terrain, closely integrated with AI [gf]22ef[/gf][gf]22ef[/gf] Ten trends of humanoid robots released

2024-08-23

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

From August 21 to August 25, the 2024 WRC (2024 World Robot Conference) was held in Beijing, and an exploration of "technology changing life" was launched. The reporter of "Daily Economic News" felt at the scene that the audience of the conference was enthusiastic, and many exhibition halls opened the "follow the crowd" mode. A large number of visitors gathered around the highly realistic "Su Shi", the robot dog that can shake hands, the humanoid robot that can play "Wing Chun", and the "pancake-making robot" that does not need to rest. It is reported that there are 169 companies and more than 600 innovative products participating in the exhibition, including more than 60 new products launched for the first time and 27 humanoid robots.

Image source: Photo by Yang Hui, reporter of China Business Network

Focusing on the development prospects of humanoid robots, at the conference forum, Qiao Hong, Chairman of the World Robotics Cooperation Organization and Academician of the Chinese Academy of Sciences, released the "Top Ten Trends in Humanoid Robots" (hereinafter referred to as "Outlook") on behalf of the organizers, analyzing ten directions including artificial intelligence-enabled humanoid robot design, humanoid robot multimodal large models, and humanoid robot embodied intelligence.

Image source: Photo by Yang Hui, reporter of China Business Network

Specifically, the Outlook points out that from the perspective of component configuration and design, in the future, core components such as high-explosive motors, high-computing power chips, precision reducers, high-precision sensors, and long-life batteries will build a more stable and high-performance humanoid robot hardware system; artificial intelligence technologies such as neural networks, graph grammar, and evolutionary algorithms will be able to automatically build the legs, feet, arms, torso and other modules of humanoid robots according to scene and task requirements, and achieve coordinated optimization of morphology and control.

In terms of motion intelligence, future humanoid robots are expected to adapt to more complex terrains and narrow environments such as slopes, stairs, and thresholds; when the lower body shakes, they will use their arms to collaborate and use human tools and equipment to complete high-performance operating tasks.

Image source: Photo by Yang Hui, reporter of China Business Network

In terms of AI interaction, the Outlook believes that the large multimodal model of humanoid robots will further enhance the robot's generalization ability in complex scenario tasks by integrating multimodal information such as voice, images, text, sensor signals, 3D point clouds, etc.; building a large-scale, standardized humanoid robot data set will be conducive to improving the humanoid robot's body design, simulation training and algorithm migration capabilities.

In addition, the Outlook also mentioned that in the future, embodied intelligence is neither a simple computer simulation in a virtual environment nor an electromechanical system that is completely biased towards physical space, but is closely related to humanoid robot systems. Humanoid robots inspired by the structure and neural mechanisms of the human body will simulate the human musculoskeletal system and neural mechanisms from the inside out, and are expected to build an efficient and stable system that is closer to humans.

Daily Economic News

Report/Feedback