A question without answer hovers over the robotics sector: what will be the consequences when the humanoid robot boom in China disappears? The warning originates neither from market skeptics nor from the country of China but from the National Development and Reform Commission (NDRC) whose spokesperson Li Chao said, “frontier industries have long grappled with the challenge of balancing the speed of growth against the risk of bubbles an issue now confronting the humanoid robot sector as well.”
Within a short period of time, the number of Chinese companies working on human-like robots has gone up to more than 150 only in China as compared to approximately 100 globally. A large number of these companies are “highly similar” models in their product models, and this overabundance may end up saturating the market with only few genuine R&D projects. The government worries about it because it was over the same route: the overinvestment in bike-sharing and semiconductors led to the harsh shakeout, thus eliminating smaller players and causing the innovation to slow down.
The real facts behind the engineering aspects of the hype should be of greatest concern to investors and analysts. Mass production cannot be talked about yet simply because two of the most difficult issues in robotics, i.e. dexterous hand design and whole-body control, are still far off being solved. The long term research of the articulated, humanlike robot hand has only resulted in the prototype for the demonstration which hardly can be generalized. Industrial grippers still use to a great extent the parallel-jaw designs or suction cup end effectors, thus being very efficient in doing repetitive, structured tasks but less efficient in unstructured environments. To achieve human-level manipulation will require strong, multi-joint fingers with tactile sensing, force feedback, and fine motor control capabilities – the features that most current humanoids hardly if at all effectively integrate.
Whole-body control is just as hard. To keep human-like bipedal locomotion in human spaces requires very complex algorithms that co-ordinate balance, gait generation, and multi-contact motion planning. The employment of ZMP-based control along with high-power joint actuation has indeed moved the field forward but the challenge to scale them for safe, autonomous, operation in neat, yet dynamic, spaces remains. Even the most advanced machines like Boston Dynamics’ Atlas or Tesla’s Optimus are reliant on controlled test environments to demonstrate their abilities.
The Chinese government is taking steps to speed up the development of the core technology, setting up training and testing facilities, and encouraging consolidation and resource sharing in the sector. Citigroup sees “exponential” growth of the Chinese program of humanoid robots next year with distant implications that project a $7 trillion market and 650 million units globally by 2050. However, the large-scale use of the robots at home or in factories will probably take several years due to the difficulties of the engineering and the problem of high prices.
The risk of a bubble has consequences for the whole world. When this bubble burst, Chinese companies that are at present aggressively pursuing production scale and cost reduction may be confronted with a shortage of capital and a wave of merger activities that in turn will delay the rollout of affordable humanoids worldwide. U.S. companies, on their part, can take a short-term advantage of reduced competition and cheaper access to Chinese components and talents but such advantages will fade out as Chinese players reorganize.
The supply-chain factors complicate the situation further. As Standard Bots CEO Evan Beard told the U.S. lawmakers, “today U.S. quotes are ten times higher than Chinese suppliers,” which points out the cost difference in the sourcing of robotic parts. Without a funded national strategy for advanced manufacturing and robotics, he warned that America would slip further behind. This need for Chinese parts – from actuators to sensors – implies that any disruption in China’s robotics industry will have effects on global production schedules and pricing.
For the hardcore fans of robotics who closely follow the technical progress, the current massive inflow of money is opening up new possibilities in experiments with AI-based manipulation learning. The likes of Figure and Tesla are investing in the vision-only methods with which the robot is supposed to learn by watching human activities through the camera. Although this method is free of the expensive motion-capture requirement, it is at the same time avoiding the essential components such as force feedback and tactile sensing which consequently limit the robot’s dexterity. Benjie Holson’s “humanoid robot Olympics” challenges like hanging a shirt with only one sleeve inside-out or cleaning peanut butter off a hand makes the point that human-level manipulation is a multidimensional problem, and visual imitation alone cannot solve it.
On top of that, China’s humanoid robot boom is a risky move due to the confluence of factors: the oversaturated competition, the unresolved engineering challenges, and the geopolitical supply-chain dependencies. For tech-savvy investors and industry analysts the main question is not only if the bubble will burst but also how its shockwaves could change the pace, the cost and the technical trajectory of the humanoid robotics across the globe.”
