You might not recall it, but you probably had a few spills as a toddler. You weren’t the only one—falling is a normal part of learning to crawl, walk, climb, and jump. During early childhood, our balance, coordination, and motor skills are still developing.
These abilities—often called physical intelligence—quickly become second nature for most people, even for surprisingly complex actions like walking, picking up objects, or moving around a room without conscious thought.
“As humans, we tend to overlook our physical intelligence because it becomes so automatic early in life,” said Bowen Weng, roboticist and assistant professor of computer science at Iowa State University.
“In reality, it’s extraordinary. The fact that you’ve managed not to fall much since age three is amazing. Once we master a skill, the physical intelligence required to execute it operates in the background, allowing our minds to focus on other things.”
Humanoid Robot Physical Intelligence Challenges
For humanoid robots, developing physical intelligence is far from automatic or straightforward.
“Despite major strides in AI, the design and adaptability of robotic bodies remain major obstacles to effective performance in real-world settings,” Weng explained.
“Robots face challenges in physical intelligence because it involves adjusting to unpredictable environments, processing sensory feedback in real time, and mastering intricate motor skills—tasks that come naturally to humans but are extremely difficult for machines to replicate.”
To address this, Weng and a team of computer science graduate students are exploring and testing new approaches to enhance the physical intelligence and capabilities of humanoid robots, while also emphasizing safety standards to ensure secure interactions between robots and humans.
The Difficulty of Adjusting
Humanoid robots are engineered to mimic the human body and interact seamlessly with human environments and tools. They serve a range of purposes, such as assisting people with tasks, supporting research, or taking on dangerous or repetitive jobs.
“To advance society and humanity, we need to accomplish more, work faster, and do it more efficiently,” Weng explained. “Automation can help us achieve that, but it’s also crucial to recognize people’s concerns and clarify that humanoid robots cannot—and will not—replace humans.”
According to Weng, the real challenge lies not in replacement, but in adaptation.
“Humanoid robots depend on humans for their design, training, oversight, ethical direction, and emotional understanding,” he said. “They cannot perform these functions independently. Therefore, our ability to adapt and strategically use AI to enhance our own capabilities is essential.”
Weng also highlighted that as humans and AI work together, new career opportunities will arise in fields such as AI supervision, ethics, design, and maintenance.
“These roles will offer fresh ways to expand both our workforce and our economy,” he added.
Cooperation Over Rivalry
At the end of a winding hallway on the first floor of Atanasoff Hall at Iowa State University, a modest door secured by a sturdy keypad hides a world of innovation: the ISU computer science robotics lab.
Inside, the lab houses two advanced legged humanoid robots—one standing about 6 feet tall, the other roughly the height of a 10-year-old—as well as a dog-like quadruped robot built to resemble a beagle.
Weng and his student team spend much of their time here, using computer software and video game–style controllers to enhance the robots’ abilities. They guide humanoid robots through actions like standing, sitting, walking, turning, and waving, while directing quadrupeds to stand, sit, leap, shake hands, and move in all directions.
“Tasks that come naturally to humans are often difficult for humanoid robots,” explained Zaid Mahboob, a computer science doctoral student. “My goal is to enable these robots to perform more of these tasks with higher efficiency.”
Mahboob emphasized that the team’s aim isn’t to pit humans against humanoid robots, but rather to enhance the robots’ precision, accuracy, and speed so they can work safely and effectively alongside humans.
Prioritizing Safety in Leadership
Weng co-authored “Repeatable and Reliable Efforts of Accelerated Risk Assessment in Robot Testing” (arXiv) and presented it at ICRA 2025. The research improved accelerated testing for robots by introducing a reliable, repeatable algorithm and assessing instability from frontal impacts.
“Watching robots operate is thrilling,” Weng remarked. “But far too little attention is being given to achieving these advances safely and responsibly.”
His study showed the algorithm performs reliably and holds promise for testing other robot types. Nevertheless, Weng emphasized that additional safety precautions are still necessary.
“This research has an ‘indirect’ effect on improving robot safety,” Weng explained. “We didn’t directly enhance the robots’ specific abilities. Instead, we improved the test algorithms’ capabilities, which could potentially be applied to robots in the future.”
Quadruped Robot Stability in Dynamic Environments
In a related study, Weng and colleagues found that commercial quadruped robots, though capable, struggle with precise body positioning on rough terrain. The study was published in the International Journal of Intelligent Robotics and Applications.
Weng noted that legged mobile robots are increasingly important for their versatility and adaptability. He noted these systems have strong potential in search-and-rescue, healthcare, manufacturing, disaster response, and mobility solutions.
Weng stated that the new research initiative aims to enhance the reliability and performance of legged robotic systems by improving evaluation standards and providing clearer insight into both their strengths and limitations.
“Transparency is essential for building public trust, ensuring safety, and enabling informed conversations about the responsible deployment and use of robots,” Weng added.
Staying Authentic
Graduate students often cite the chance to work with real robots as a key reason for joining Weng at Iowa State.
“Most robotics labs focus primarily on simulations, but in Iowa State’s lab, we carry out real experiments with real robots to collect real data,” said Yuija Chen, a doctoral student in computer science.
Dylan Khor, a second-year master’s student in computer science, also cited the chance to collaborate closely with Weng as a key motivator.
“I first took Dr. Weng’s introduction to machine learning class and really enjoyed his teaching style,” Khor said. “He brought immense energy and expertise to the course, and it was very interactive. So when I decided I wanted to pursue robotics research, I reached out to him immediately. It’s been an incredible experience.”
While Weng focuses on standardizing robot safety, he also finds great fulfillment and inspiration in mentoring students. Beyond collaborating with graduate students, he oversees several undergraduate computer science teams working on programming legged robots and robotic arms.
Weng said research opportunities will grow once the new robotics lab in Durham Hall opens. Set to finish this academic year, the lab already has eight robotic arms for teaching and research.
Prospects for Adoption
Weng noted that humanoid robots face hurdles like high costs, lack of standards, limited infrastructure, and few practical applications. Additionally, social and ethical issues must be addressed.
Nevertheless, Weng emphasized that these challenges make continued research into improving robot safety and effectiveness even more critical.
“Ultimately, trust is essential,” Weng said. “The way to establish the trustworthiness of humanoid robots is through research guided by humans.”
Read the original article on: Tech Xplore
Read more: What is the Domain Name System? An Engineer Explains