Workflow
宇树科技王兴兴:通用人形机器人最早2026年底实现“自主干活”

Group 1 - The core viewpoint of the article is the evolution of embodied intelligence and humanoid robots, highlighting the clear roadmap for development and the current technological challenges faced by the industry [1][2]. - The development of general humanoid robots is segmented into key milestones, with the first milestone achieved in 2023, allowing robots to understand complex commands and perform smooth actions [1]. - The next critical milestone is expected to be reached by the end of 2023 or mid-2024, enabling robots to respond to real-time commands and perform a range of actions within their capabilities [2]. Group 2 - A significant breakthrough would be enabling robots to autonomously understand commands and interact with their environment in unfamiliar settings, potentially achievable by the end of 2024 or 2025 [2]. - The long-term goal is to increase the task success rate of robots to over 99.9% and enable them to perform intricate tasks like disassembling a smartphone, which may take several more years [2]. - Key challenges include the power consumption and computational requirements of the robots, with an ideal target of keeping terminal power consumption below 100 watts, and average operational power around 20 to 30 watts [2]. Group 3 - The reliability of communication and cabling in robots is crucial, with cable issues accounting for over 60% of industrial robot failures, emphasizing the need for improved communication protocols and reduced internal wiring complexity [3]. - The industry is currently in a "dawn" phase before a potential explosion of growth in embodied intelligence, with a call for collaboration among chip manufacturers, OS developers, and algorithm companies [3]. - The company has open-sourced a self-developed vision-related model along with datasets and training codes to encourage broader participation in solving industry challenges [3].