Watch: Robots smash grueling 3,100-step mountain climb in a world first

While robot dogs feel like old news now, setting them the challenge of racing up more than 3,000 uneven stone steps on the side of a mountain is more than just to show off what machines with legs can do. It’s a huge lesson in their training.

On November 2, robot dogs – and other kinds of machines built for locomotion of some kind – took part in the very first World Robot Mount Taishan Climbing Competition in Tai’an, in China’s Shandong province.

According to a report in Global Times, nearly 90 robot-making teams flexed their engineered legs in yet another test of their abilities. We’ve seen robots succeed and fail over sprints and marathons, but conquering mountains is a new challenge.

You can see the highlights courtesy of footage courtesy of CCTV+ media.

“Humanoid robots and robot dogs tested their climbing capabilities,” the Times described. ” They were seen climbing, pausing, recalculating, and then moving forward again with surprising agility. Some moved with careful precision, others with a quick, springy gait, giving spectators a glimpse of how far robotics has come, and how far it still has to go in real-world scenarios.”

The route – 3.9 km, or just over 2.4 miles, with an elevation of around 600 m (0.37 miles) – would take an average hiker an hour or two. The best bots did it in less than 60 minutes. But like humans, not all machine climbers conquered.

The one that did, however, was a robot from Beijing’s Lingyi Technology, a subsidiary of Hangzhou’s famous Unitree Robotics. The four-legged robot scored top marks for speed, finishing the route in just 43 minutes and 41 seconds.

Interestingly, in trial runs of the course – from Hongmen to Zhongtianmen – rival Beijing company UNIUBI AI put itself on the map, with its flagship Cyvet robot dog clocking a time of 35 minutes and 47 seconds. Isn’t that always the way, getting a PB time when it’s just a trial or warm-up run …

“Adapting to Mount Taishan’s road conditions was the biggest challenge,” Chen Lin, a navigation engineer on the Hangzhou-based UNIUBI AI team told state media CCTV. “These steps were built centuries ago. Some are high, some low, some wide, some narrow, and there are also cobblestone sections.”

According to reports, the smallest robot in the competition, Cyvet used LiDAR and real-time 3D mapping to master the uneven terrain. So, sort of like what a human brain does as it navigates such novel challenges.

Mount Taishan is the tallest peak in the province, and while there are several routes for hikers, some more challenging than others, and the longest climb features more than 7,000 steps. However, no route is particularly easy and you can expect to be spending several hours on steps even taking the “simplest” path.

The competition’s four official events featured humanoid robots racing to the top, robots load-carrying up the mountain, an exoskeleton-assisted machine climbing, and an all-terrain scouting challenge.

Ultimately, though, the event specifics weren’t the focus; they were more ways to test the advancement of these two- and four-legged robots that have so far done most of their movement on pretty flat earth.

Climbing precarious and unpredictable steps really tests some important mechanics as the technology develops – including balance, real-time decision-making and adaptation. These attributes will be increasingly important when robots are deployed to rescue sites and even used in less dangerous settings like hospitality and in multilevel offices, factories and houses.

“Like the early exhibitions of aircraft and automobiles, robot sporting events can serve as visible indicators of technological maturity, attracting public attention while also drawing investment,” Xiang Ligang, president of the Beijing Institute of Intelligent Era Industry, told the Times.

It’s easy to think such events seem like a waste of time and resources – especially when they come with videos of robots falling over and falling apart. But like earlier athletic tests, the climb showed where engineers need to improve to compete in this market.

Balance – especially for taller bipedal humanoids – will be one of the biggest hurdles for these kinds of robots, as will real-time decision-making without a human minder in tow, if we can expect to see them as reliable consumer products anytime soon. And events like this are designed to speed up the process beyond the lab.

Also, it’s worth pointing out that these bots, even the ones not at the front of the pack, deserve some real credit.

If you’ve climbed anywhere in China, you’ll know that even with clearly defined stone steps, it’s not like taking a few sets out of a subway station; they’re sturdy but often uneven, many are centuries old, and handrails are a pleasant surprise but unexpected. And there’s a reason why escalators and elevators have become popular for getting humans up mountains in recent years. There are even conveyor belts to ferry folks up steep inclines – which also makes some of China’s most beautiful views easier to reach for those less physically able.

Meanwhile, if you prefer the idea of going down a mountain in style, you can now take part in real-life Mario Kart – without the pranking of other drivers, of course – down Gulong Gorge, a 3.7-km (2.3-mile) winding track that gets you from the top to the bottom at a top speed of 30 km/h (18 mph).

For now, we’re yet to see robots master this at least.

However, the Times reported that China’s Ministry of Industry and Information Technology listed more than 400 different humanoid models in production and had rolled out numbers exceeding 40,000 units in the first half of 2026.

Source: Global Times