一袋零食，难倒了11支顶尖机器人团队

Core Insights - The recent AgiBot World Challenge at IROS 2025 highlighted the importance of not only algorithms but also the engineering deployment capabilities of teams in robotics [1][10] - All participating teams failed to complete the task of hanging snacks, indicating significant challenges in visual recognition and spatial positioning for robots [2][4] Group 1: Task Performance - The task of hanging snacks was expected to be straightforward, yet no team successfully completed the entire sequence of actions [2] - The highest-scoring team only managed to score points for grabbing the snack bag, demonstrating the difficulty of the task [2] - In contrast, teams performed well in tasks like folding clothes and pouring water, suggesting advancements in pre-training and imitation learning for medium complexity tasks [5] Group 2: Challenges in Robotics - The competition revealed that the real challenge in robotics has shifted from executing actions to understanding the environment, particularly in dealing with visual noise and spatial interactions [9] - The complexity of the task was exacerbated by environmental factors, such as the difficulty in distinguishing snack packaging due to color similarity and the small size of hooks [4] Group 3: Industry Implications - The event served as a platform for validating the entire chain from data collection to model deployment, emphasizing the need for robust engineering processes in industrial applications [10] - The introduction of the "LingChuang" platform aims to lower barriers for secondary development, facilitating broader participation in robotics development [10] - The outcomes of the competition reflect a critical intersection of algorithmic advancements and engineering capabilities, which are essential for the practical deployment of general-purpose robots [10]