机器人大讲堂·2026-03-19 04:08Core Viewpoint - The collaboration between Qianxun Intelligent and JD.com marks a significant advancement in the application of embodied intelligence in retail, transforming data collection and model training into a practical, integrated process that addresses the industry's data scarcity issue [1][5][25]. Group 1: Strategic Collaboration - Qianxun Intelligent and JD.com signed a strategic cooperation agreement from 2026 to 2029, focusing on customized consumer products, technology collaboration, and marketing co-construction to accelerate the application of embodied intelligence in retail [3]. - This partnership establishes a "technology + scene + data" model, where JD.com evolves from a retail channel to a data partner, and Qianxun Intelligent integrates deeply with JD.com to streamline the entire process from data collection to commercial implementation [5][6]. Group 2: Importance of Retail Scenarios - Retail environments are crucial for the evolution of embodied intelligence models due to their ability to generate high-value real-world interaction data, which is more valuable than traditional internet-based data [6]. - The collaboration allows for the collection of expert-level data through real-world operations, enhancing the model's training and adaptability [10][16]. Group 3: Data Collection and Model Training - Qianxun Intelligent has accumulated over 200,000 hours of real interaction data, with plans to exceed 1 million hours by 2026, significantly aided by the partnership with JD.com [13][15]. - The company emphasizes the importance of "dirty data," arguing that real-world imperfections provide better training scenarios for models, enhancing their generalization capabilities [15]. Group 4: Data Loop and Model Iteration - The data collection process is structured into three phases: initial remote operation for standardization, supervised robot task completion for complex scenarios, and full autonomous operation for a complete data loop [17][18]. - This iterative process ensures that data collection aligns with actual application needs, avoiding inefficiencies and enhancing the commercial viability of the technology [18][24]. Group 5: Cross-Industry Application - The data loop model developed by Qianxun Intelligent demonstrates strong adaptability across different sectors, including industrial and retail, proving its practical value and enabling broader applications [19][24]. - Future plans include extending robotic applications to JD.com’s pharmacy for automated sorting and precise medication dispensing, as well as other retail scenarios [24]. Group 6: Future Outlook - The year 2026 is anticipated to be pivotal for the commercialization of embodied intelligence, with the collaboration between Qianxun Intelligent and JD.com expected to drive significant advancements in the industry [25][26].