AI news ·
Robots complete first climbing competition on Mount Tai
A robot climbed 3,100 steps up Mount Tai in 43 minutes, 41 seconds to win China's first robot mountain-climbing race, beating more than 50 teams over a 3.9-kilometer course.
A robot climbed 3,100 stone steps up Mount Tai in under 44 minutes on Monday, winning the first China Media Group World Robot Climbing Mount Tai Competition in Tai'an, Shandong province. More than 50 embodied AI companies, universities, and research teams entered the 3.9-kilometer course from Hongmen to Zhongtianmen, with a cumulative elevation gain of nearly 600 meters. Organizers billed it as the world's first professional robotics competition held in a mountainous UNESCO World Heritage Site.
The course was a working test, not a demo. Steps varied in height and width, surfaces were uneven, and gradients shifted without warning. Organizers described the trail as a natural testing ground for pushing embodied AI systems to their limits.
Who won what
The legged robot race went to Beijing Lingyi Technology, a Unitree Robotics subsidiary, in 43 minutes and 41 seconds. Shanghai Gaoyi Technology, also running Unitree robots, took second, and the Lingju team from Linxai Intelligent finished third.
Endurance decided the load-bearing event. The Lingshi team from Linxai Intelligence won in one hour, 51 minutes, and 29 seconds, ahead of Shanghai Gaoyi Technology and Shenzhen Technology University. The humanoid race was shortened to the stretch from Hongmen Square to Doumu Palace because of the terrain; the Zhishen Lingyue team won in 27 minutes and 26 seconds. A separate all-terrain mobility and pathfinding challenge ran alongside the four main events, and China Media Group broadcast the whole thing across its platforms.
Where the machines struggled
Laboratory floors are flat and controlled. Mount Tai isn't. Uneven steps demanded more precise terrain perception and gait planning, while the continuous climb tested dynamic balance and joint control. Long runs at high intensity put pressure on motor cooling, battery life, and system stability.
Some robots overheated their motors, dropped into automatic power reduction, and drifted sideways. Several load-bearing robots developed unstable gaits after picking up new payloads at supply points. None of that ended a run outright, but it maps the engineering gap between walking on a flat surface and working outdoors. Robots also had to spot obstacles on their own and recover after losing balance or falling.
The event drew teams working on control and perception problems that show up well beyond a mountain trail. Programs covering AI R&D engineering track the same shift from simulation to field testing.
More than a race
The competition ran alongside the Mount Tai Embodied Intelligence Industry Summit, a robotics industry chain exhibition, and an investment matchmaking event. Companies, research teams, and investors used the three days to compare products and look for partners.
The trail itself produced data: continuous uphill climbs, complex terrain, and extended operating periods that are hard to reproduce in a lab. Organizers say the results could guide product development and inform use of humanoid and legged robots in inspection work, mountain rescue, and outdoor transport.
Why this matters for hospitality and events teams
Venues and event organizers are watching the same question this race answered in public: can a machine hold up for hours on unfamiliar ground with an audience present? The failures on Mount Tai - overheating, power throttling, unstable loads - are the ones that would end a demo at a trade show or a resort property. If you're scouting robots for guest services, logistics, or outdoor activations, ask vendors for field endurance data and payload-change behavior, not lab footage. The research on embodied AI is moving from controlled tests to real courses, and procurement questions should follow it.