一袋零食，难倒了11支顶尖机器人团队

Core Viewpoint - The recent AgiBot World Challenge at IROS 2025 highlighted the challenges in robotic task completion, particularly in the "hanging snacks" task, where no team succeeded in completing the full sequence of actions, indicating significant hurdles in visual recognition and spatial positioning for robots [3][4][10]. Group 1: Task Performance - All participating teams failed to complete the "hanging snacks" task, which was surprising given the expectations [3][4]. - The highest-performing team only managed to score points for the action of grabbing the snack bag, not for the complete task [5]. - The complexity of the task was exacerbated by environmental factors, such as visual noise and the small size of the hooks, which hindered the robots' ability to accurately identify and interact with the target [7][8]. Group 2: Technological Insights - The competition revealed that while robots have made progress in executing actions, the real challenge lies in understanding the environment, particularly in dealing with visual noise and physical interactions [10]. - The success in simpler tasks, like folding clothes and pouring water, suggests that teams have developed effective pre-training and imitation learning strategies for moderately complex operations [8][10]. Group 3: Industry Implications - The event served as a platform for validating different research paths and highlighted the gap between theoretical model performance and practical industrial application [11]. - The release of the new "LingChuang" development platform aims to lower the barriers for secondary development, facilitating broader participation in robotic development [12]. - The ongoing advancements in algorithms and the establishment of comprehensive development tools are critical for the successful deployment of general-purpose robots in real-world scenarios [12].