燃爆科技圈！ICRA2026顶会REAL-I具身智能挑战赛，诚邀全球算法高手巅峰对决！

Core Viewpoint - The IEEE ICRA 2026 has officially launched the first Real-world Embodied AI Learning Challenge (REAL-I), aimed at advancing research in embodied intelligence and data-driven robotic manipulation, focusing on real industrial tasks as evaluation criteria [1][5]. Group 1: Competition Overview - The REAL-I challenge is organized by IEEE ICRA, Lejoin Intelligence, and Beijing Institute of General Artificial Intelligence, in collaboration with top global universities [1][35]. - The competition is designed to test algorithms on dynamic grasping and dual-arm coordination, which are critical technical challenges in robotics [1]. - Successful algorithms validated in the REAL-I challenge will be applicable in logistics and manufacturing industries, positioning participants to gain a competitive edge in the "embodied intelligence" sector [1]. Group 2: Competition Structure - The competition consists of two phases: a qualifying Simulation Track followed by a Real-Robot Stage, with each task scored out of 100 points [16][27]. - The Simulation Track allows participants to validate and refine their algorithms for stability and reliability before advancing to the Real-Robot Stage [27][28]. - The final evaluation will be conducted on the KUAVO-4 Pro full-size humanoid robot [28]. Group 3: Tasks and Evaluation - Task 1 involves weighing express parcels accurately [17][18]. - Task 2 focuses on sorting parts from a conveyor belt into the correct bins [21]. - Task 3 requires full-cycle plate transporting, including moving to a target area, picking up a tray, and placing it successfully [23]. Group 4: Prize Pool and Recognition - The total prize pool for the competition is $90,000, with the champion receiving a KUAVO-4 Pro robot valued at $70,000, and additional prizes for runners-up [3][31]. - There is also a $500 reward for each of the top 10 published papers based on the competition platform and datasets [33].