First CMG World Robot Mount Tai Competition held; experts discuss industrial deployment of complex-terrain robots

Robots climbing Mount Tai, what exactly is being competed? On November 2, the first CMG World Robot Climbing Mount Tai Competition was held in Tai'an, Shandong. Nearly 90 teams engaged in real-scene competition on 3,100 steps, focusing on testing the all-terrain passage capabilities of various products. From competition venues and exhibition halls to mine mining, high-altitude operations, and emergency rescue, robots are accelerating toward real-world scenarios. What kind of hard power do they actually possess? How far is it from large-scale deployment? Which industries have the most urgent application needs? “News 1+1” connected with Professor Zhang Guoteng from the School of Control Science and Engineering at Shandong University and Senior Engineer Wang Xiao from the Intelligent Research Institute of the China Industrial Internet Research Institute to look at the inner workings through the competition venue.

Autoplay

△Video | “News 1+1” Full Version

Why let robots compete in mountain climbing?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: Compared with flat ground and laboratory scenes, climbing Mount Tai can provide more continuous unstructured terrain, such as steps and uneven slopes. We could, of course, also build similar scenes in laboratory settings, but it is difficult to have continuous, long-distance scenes like Mount Tai. For robots, first, it is necessary to test their adaptability under different terrains, and also to test the reliability and stability of the system in extreme environments like Mount Tai.

What problems occurred during the robot's ascent of Mount Tai that did not appear in the laboratory?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: First, during the process of the humanoid robot continuously running and going up and down steps on the winding path of Mount Tai, the screws on the feet broke off directly. This situation of screw breakage had never occurred in the laboratory. Additionally, the different terrains and temperature environments on the mountain caused the power consumption to be much higher than we predicted. Mechanical structure damage and energy loss are things that are difficult to encounter in a laboratory environment.

What kind of changes does the fact that the robot dog runs too fast to prompt an adjustment in the final competition format indicate?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: This indeed reflects the development of technology. Robot technology has advanced too quickly in these last one or two years; previously, we were thinking about how robots could walk more steadily, but this time, robots are already able to sprint on mountainous terrain and carry very heavy loads. Like the champion of this quadruped robot competition, the time taken is already shorter than that of the vast majority of human contestants.

In these few years, our robots have not made progress on single points, but rather progress in the entire system. In terms of the body, the motor driving power is stronger, and the body is more reliable and lighter; in terms of the robot's “little brain,” motion control is more flexible; in terms of the “brain,” capabilities such as autonomous decision-making and environment perception have also been greatly improved. The technical progress reflected in the competition is very exciting.

How will the capabilities of robots be applied to complex environments in the future?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: A speed race compares speed, but the ultimate goal is not speed, but whether robots can quickly, stably, and autonomously reach a specified destination and complete target tasks under various complex terrains. In the future, robots may be used for mountain patrol, disaster rescue, including fire scene reconnaissance, etc., and their speed and timeliness are crucial.

From an application perspective, quadruped robots are currently developing faster and earlier, and their applications are also relatively mature; they already have good applications in industrial inspection, logistics inspection, including military, and will further expand application scenarios in the future.

Another speed competition scenario this time—humanoid robots—is currently in a stage of rapid iteration and development from research to application. In fields familiar to everyone such as education, scientific research, performance, and entertainment, humanoid robots already have applications; in fields such as automobile manufacturing and industrial logistics, humanoid robots have also begun application pilots now.

What did you and your team members gain through this competition?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: The competition provided us with a great opportunity for testing. In this competition, Mount Tai provided us with a great multi-terrain environment, including continuous, irregular steps, and rugged mountain roads. In this process, we obtained joint torque data, trunk posture, and energy consumption data, etc., These are difficult for us to obtain in the laboratory, and they have great guidance and reference significance for our future research.

Additionally, we encountered various unexpected problems on the mountain, including mechanical structure, energy consumption, as well as problems in perception and control. These problems allowed us to accumulate more engineering experience. I believe that this kind of engineering data and engineering experience on the mountains is a more important gain in research than the results.

Will robots take away the jobs of the mountain porters?

Autoplay

Professor Zhang Guoteng of the School of Control Science and Engineering, Shandong University: Mountain porters have already become a cultural symbol of Mount Tai, but they are currently working very hard, and they are also facing a labor shortage. In the future, the development of bipedal robots can reach a state of human-robot collaboration; we can use robots to alleviate the pressure on mountain porters, and mountain porters can also operate robots through some simple training to help them carry heavier objects and complete material transportation tasks over longer and higher distances. This is also a reflection of technology changing life and improving the quality of life.

To which scenarios is the capability presented by the robots in this competition most likely to be promoted first?

Autoplay

Wang Xiao, Senior Engineer at the Intelligent Research Institute of the China Industrial Internet Research Institute: This robot mountain climbing competition can be described as a comprehensive stress test under complex environments. The achievements the robot has achieved now indicate that its mobility is basically up to standard.

First, this competition reflects the hard skills of the robot's continuous passage. For example, in the legged racing competition, the robot must continuously complete 3.9 kilometers and over 3100 steps; this involves technical points such as terrain perception, foot placement planning, whole-body balance, and energy consumption control, where any technical shortcoming would cause the robot to stop.

Secondly, this competition's balanced rules included the weight of autonomous operation in the overall score. This indicates that what we are comparing is not who runs faster, but who can still run steadily under minimal human intervention, which is a more meaningful direction for the entire industry.

Third, it involves the boundaries of capability in the competition. The competition is actually a “capability physical exam,” but it is not a “job qualification certificate.” Mountaineering mainly tests mobility and terrain adaptability, but the actual work positions in the industry, such as industrial positions, also involve action segments such as grasping, plugging in, and inspection; we must be responsible for quality, pace, and safety, and this difficulty is not on the same level as a mountaineering competition.

Our current capabilities can be migrated first, but they have not yet reached the level of replication. On the basis of mountain climbing, we can carry out large-scale promotion to inspection scenarios such as power, tunnels, and chemical industrial parks. But to truly be competent for real jobs, it still needs to undergo tests of long-term operation and economic feasibility. From walking well and standing well to working well all the time, this is the final result we want to achieve.

How should the priority of replacing manual labor with robots in industrial scenarios be arranged?

Autoplay

Wang Xiao, Senior Engineer at the Intelligent Research Institute of the China Industrial Internet Research Institute: Not all industrial scenarios require humanoid robots, or require a certain type of robot, but a large number of industrial tasks indeed require the capability of embodied intelligence. My criterion lies in whether this position should actually be done by a robot. Here we focus on three aspects:

First, the demand side must be large enough. We conducted a systematic review of the entire industrial system in the early stage and identified over 900 feasible scenarios, of which those with high application value and strong feasibility account for around 12%.

Second, the most urgent demands are for hazardous positions. In fields such as mining, energy, metallurgy, and chemicals, there are some high-temperature, toxic, dusty, and other hazardous work scenarios. These scenarios face problems of difficulty in recruiting and high risks, where the demand for replacement is most urgent. We can start by entering these scenarios first, for example, starting from inspection, and then extending to handling, transferring, and simple operations, to carry out large-scale promotion.

Third, those that can appropriately slow down the pace are high-precision, high-tempo positions. For example, precision assembly in industry, flexible material operation, as well as low-tolerance workstations. These scenarios are currently not suitable for hurried promotion; their requirements and difficulties are relatively high, and they are suitable as medium-to-long-term breakthrough directions.

So our promotion sequence is still First solve the places where people cannot go, then solve the places where people are unwilling to go, and finally the places where people do not need to go.

In high-risk scenarios of “dangerous, dirty, tiring, repetitive, and difficult,” to what extent can the current level of robots complete tasks?

Autoplay

Wang Xiao, Senior Engineer at the Intelligent Research Institute of the China Industrial Internet Research Institute: From the current capabilities of robots, to complete these high-risk scenarios, at least four more

First is the “body gate,” which means the robot body must possess safety protection capabilities such as explosion-proof and dust-proof.

The second stage is the “Operation Stage,” where we can make it perform work that is accurate in seeing, steady in gripping, and capable of twisting.

The third stage is the “Intelligence Stage,” which is the autonomous handling capability for abnormal situations.

The fourth level is “Safety Baseline,” allowing the robot to stop and seek help when it exceeds boundaries, or to be remotely taken over by a human.

Producer | Wang Xiaochen

Planner | Wang Qian

Director | Guo Dandan, Bi Haoyun, Liu Zhe, Zhao Wei

Production | Li Yang, Ji Yang, Xie Chaoquan, Lyu Yang, Duan Haochen