How researchers are working to soundly advance capabilities of humanoid robots

It’s possible you’ll not keep in mind it, however odds are you took just a few tumbles throughout your toddler period. You were not alone. Falling, in spite of everything, is a pure consequence of studying to crawl, stroll, climb and leap. Our stability, coordination and motor expertise are creating all through early childhood.

Nevertheless it additionally would not take lengthy for these skills—also referred to as bodily intelligence—to grow to be second nature for many, together with deceptively complicated actions corresponding to strolling, greedy objects and navigating our manner throughout a room with out having to consider it.

“As people, we regularly take our bodily intelligence as a right as a result of it turns into so computerized after we’re nonetheless younger,” mentioned Bowen Weng, roboticist and assistant professor of laptop science at Iowa State College.

“However the reality is, it is outstanding. The truth that you have not fallen over a lot because you had been 3 years outdated is outstanding. As soon as people grasp a talent, the bodily intelligence wanted to carry out that talent turns into a background course of, and our minds are freed as much as give attention to different duties.”

For humanoid robots, nevertheless, bodily intelligence is something however computerized or simple to be taught.

“Even with vital advances in AI, the bodily design and adaptableness of robotic our bodies are nonetheless vital boundaries to real-world efficiency,” Weng mentioned.

“Robots battle with bodily intelligence as a result of it requires adapting to unpredictable environments, integrating sensory suggestions in actual time and mastering complicated motor expertise. These duties are largely intuitive for people however notoriously tough to copy in machines.”

That is why Weng, together with a number of graduate college students in laptop science, are working to develop and check new strategies that might assist advance the bodily intelligence and capabilities of humanoid robots whereas additionally prioritizing the event of security requirements that can enable robots and people to work together safely.

The problem of adaptation

Humanoid robots are designed with a physique form that resembles the human type and are constructed to work together with human environments and instruments. They can be utilized for quite a lot of functions, together with aiding people with duties, conducting analysis, or performing harmful or tedious jobs.

“To assist transfer our society and humanity ahead, now we have to get extra achieved, do extra issues and discover a option to accomplish all of it with higher velocity and effectivity,” Weng mentioned. “Automation is a step to assist us attain that purpose, however on the identical time, it is necessary to acknowledge sure fears individuals may need and clarify why people cannot and will not get replaced by humanoid robots.”

The problem, Weng mentioned, is not alternative however adaptation.

“Humanoid robots want people as a result of they depend on us for design, coaching, supervision, moral steerage and emotional context,” he mentioned. “They can not replicate or perceive these capabilities on their very own, so our willingness to adapt and make the most of AI to strategically help and increase our personal skills is de facto necessary.”

Weng additionally famous that new profession paths and jobs will emerge within the areas of AI oversight, ethics, design and upkeep as we adapt to human-AI collaboration.

“These might be new alternatives to develop our workforce and financial system,” he mentioned.

Collaboration, not competitors

On the finish of a winding again hallway on the primary ground of Atanasoff Corridor on the Iowa State campus, there’s an unassuming door secured with a sturdy keypad lock and a little bit of thriller. On the opposite facet is an area full of cutting-edge potentialities: the ISU laptop science robotics lab.

At present, the lab is dwelling to 2 extremely superior legged humanoid robots, with the most important standing about 6 ft tall and the smaller reaching roughly the peak of a 10-year-old baby, in addition to a quadruped robotic with dog-like attributes and the construct of a beagle.

It is also the place Weng and his crew of scholar researchers can usually be discovered working with laptop software program and hand-held controllers, very similar to these used with video-gaming techniques, to assist advance the bodily capabilities of those robots. This consists of taking the humanoid robots via instructions to face up from a mendacity place, sit down, stroll, flip round, wave their arms and shake palms, and directing the dog-like quadruped robotic to face, sit, leap, shake palms, and stroll ahead, backward and sideways on command.

“Issues that people are capable of do simply, a humanoid robotic shouldn’t be in a position to try this,” mentioned Zaid Mahboob, a doctoral scholar in laptop science. “My motivation is assist make these robots able to doing extra of this stuff with higher effectivity.”

Mahboob mentioned the group’s shared purpose is not to create competitors between people and humanoid robots, however as an alternative to enhance the precision, accuracy and velocity of the robots to allow them to safely and successfully collaborate with people.

Main with security

Weng not too long ago co-authored the paper “Repeatable and Dependable Efforts of Accelerated Threat Evaluation in Robotic Testing” obtainable on the arXiv preprint server and introduced its findings through the 2025 IEEE Worldwide Convention on Robotics and Automation. The research aimed to reinforce present accelerated testing frameworks for robots by proposing a brand new algorithm that is each repeatable and dependable whereas additionally assessing a robotic’s danger of instability from disturbances brought on by frontal impacts.

“It is thrilling to see robots work,” Weng mentioned. “However on the identical time, there is not practically sufficient effort being put into responsibly and safely attaining these developments.”

Weng’s analysis confirmed that the brand new algorithm works reliably and with excessive chance, and that future testing with different robotic varieties is promising. Nevertheless, extra security measures are nonetheless wanted, he mentioned.

“This analysis makes an ‘oblique’ affect in direction of a safer robotic,” Weng mentioned. “We didn’t make any contribution to reinforce the robotic’s particular capabilities. We did do one thing to reinforce the check algorithm’s particular capabilities, which can be utilized towards the robots sometime.”

A second research, “Experimental Analysis of Industrial Quadruped Robots: Stability and Efficiency in Non-inertial Environments” co-authored by Weng in collaboration with researchers on the College of Massachusetts at Lowell and Purdue College, discovered that whereas the quadrupeds confirmed eager capabilities, challenges with correct physique positioning throughout aggressive floor movement stay. The work is printed within the Worldwide Journal of Clever Robotics and Functions.

“Cell robots that characteristic limbed constructions—significantly legs—have gotten more and more related due to how versatile and adaptable they are often throughout many alternative environments and functions,” mentioned Weng, noting that legged robotic techniques have proven vital promise in functions corresponding to search-and-rescue missions, well being care help, manufacturing, catastrophe response and enhanced mobility options.

Weng mentioned the brand new analysis venture will contribute to the improved reliability and effectiveness of legged robotic techniques by advancing analysis requirements and by selling transparency concerning the capabilities—and limitations—of those techniques.

“Transparency is important for fostering public belief, enhancing security and serving to us have significant discussions across the sensible deployment and accountable use of robots,” Weng mentioned.

Retaining it actual

Among the many graduate college students working with Weng, having the prospect to conduct analysis utilizing actual robots is often cited as a key promoting level of their resolution to come back to Iowa State.

“Many of the robotics labs are simply fascinated with simulation and that form of factor, however in (Iowa State’s) lab, we’re conducting actual experiments utilizing actual robots to get actual information,” mentioned Yuija Chen, a doctoral scholar in laptop science.

Dylan Khor, a second-year grasp’s scholar in laptop science, mentioned he was additionally motivated by the chance to work intently with Weng.

“I first took (Dr. Weng’s) class on introduction to machine studying, and I actually preferred the way in which he taught the category,” Khor mentioned. “He introduced a lot power and data to the topic, and the category was very interactive. So, once I determined I needed to do analysis in robotics, I reached out to him straight away and right here I’m. It has been a beautiful expertise.”

Whereas Weng’s focus is on creating standardized security measures for robots, he additionally finds nice objective, enjoyment and inspiration in serving to information college students of their analysis. Along with working with graduate college students, Weng leads a number of undergraduate laptop science groups in programming work with legged robots and robotic arms.

Weng mentioned he expects analysis alternatives will continue to grow when the brand new robotics training lab is totally up and operating within the decrease stage of Durham Corridor. This training lab, which is slated for completion through the present educational 12 months, already consists of eight robotic arms for analysis and instructing functions.

Adoption outlook

Weng mentioned humanoid robots nonetheless face some main hurdles to widespread adoption, together with excessive growth and unit prices, an absence of standardization, infrastructure gaps and restricted use instances, and there are social and moral challenges that must be addressed as properly.

Nevertheless, Weng mentioned these challenges and development boundaries additionally make ongoing analysis to enhance the protection and efficacy of robots much more necessary.

“The underside line is, you could have to have the ability to belief it,” Weng mentioned, “and the trail to proving the trustworthiness of humanoid robots is thru human-led analysis.”