New autonomous system helps stair-climbing robots recover from falls using reinforcement learning and a robotic arm.
Robots that navigate staircases still face a major safety problem as losing balance often causes them to tumble with increasing force, risking damage to themselves and their surroundings.
A new study from researchers at the Robotics and Automation Research Laboratory at the Singapore University of Technology and Design proposes a way to reduce that risk by helping robots actively mitigate falls rather than only trying to prevent them. The research focuses on stair-traversing service robots, which remain vulnerable despite advances in balance control and path planning.
According to a multi-year field study cited by the team, robots designed for stairs fail at least 35 times more often on staircases than on flat ground. Once a fall begins, momentum builds quickly, making recovery difficult and increasing the risk of damage. Traditional safety systems mainly focus on fall prevention. These include navigation algorithms, obstacle detection, and stability control.
However, the researchers argue that these measures cannot fully eliminate risk in real environments. Even if a robot is well-controlled, unexpected interactions, such as a person bumping into it, can trigger a fall that existing systems are not designed to handle.
“This is why fall mitigation matters more than fall prevention alone,” said Professor Mohan Rajesh Elara, who heads the ROAR Laboratory at SUTD. “Until the industry has a credible answer to that residual risk, operators will keep treating these platforms as a liability rather than a labour-saving tool. ” To address this gap, the team developed a reinforcement learning-based system that enables a robot to respond during a fall.
Instead of only reacting before instability occurs, the system actively attempts to stabilise the robot mid-fall using a mechanical arm attached to a tracked mobile platform. The researchers first studied how stair-related failures occur in real conditions. They identified five primary fall modes, including straight backward falls, pivoting variations, and sideways collapses. Based on this analysis, they designed a three-jointed robotic arm capable of countering all five types of motion.
“Three degrees of freedom turn out to be the minimum that can geometrically cover the five fall types, when the structure is mounted at the rear of the robot,” said Professor Elara. “The mechanism narrows the problem enough that an AI controller can solve it. The AI lets us drive a mechanism that would otherwise be too complex to control by hand. ”learning.
In each training episode, the robot was subjected to simulated external forces that pushed it toward a fall. The controller then learned how to adjust the arm joints in real time to stabilise the platform. Across five trained controllers, the system achieved an average success rate of 69.4 percent in stopping falls and restoring stability. A traditional hand-coded control method achieved 38.6 percent.
When successful, the AI system stabilised the robot in an average of 4.25 seconds, meeting the team’s internal response target of under 10 seconds.it was trained on. They evaluated it on platforms that were 10 percent larger or smaller and on staircases with different dimensions. In some cases, performance improved, with the best controller reaching an 87 percent success rate on a larger robot.
“The controller is not memorising one geometry,” said Prof Elara. “It is learning a recovery strategy that generalises. ” The team says the system is not yet ready for full deployment. Its current performance does not meet strict industrial safety standards such as IEC 61508, meaning it would need additional safeguards and validation before real-world use.
The researchers are now working on improving reliability, adding mechanical fail-safes, and developing explainable models that can support certification. The work is part of a broader effort at SUTD to improve the safety of autonomous mobile robots in real environments. With over a decade-long career in journalism, Neetika Walter has worked with The Economic Times, ANI, and Hindustan Times, covering politics, business, technology, and the clean energy sector.
Passionate about contemporary culture, books, poetry, and storytelling, she brings depth and insight to her writing. When she isn’t chasing stories, she’s likely lost in a book or enjoying the company of her dogs. AI and Robotics
Autonomous Robots Fall Mitigation Industrial Safety Reinforcement Learning Robotics Stair-Climbing Robots SUTD
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
New Jersey, New York investigate FIFA World Cup ticket practices and pricesThe attorneys general for New Jersey and New York launched an investigation into FIFA's ticketing prices and practices for the 2026 World Cup.
Read more »
New York, New Jersey attorneys general launch FIFA probe investigating World Cup ticketing policiesNew York Attorney General Letitia James and New Jersey Attorney General Jennifer Davenport have launched a joint investigation into FIFA's World Cup ticketing
Read more »
New low-voltage pea-sized pump makes soft robots lighter, more agilePea-sized liquid-metal pump could replace bulky compressors in future soft robots and wearable devices.
Read more »
Susan Boyle drops hint at new project with glamorous new look and 'new era' teaseSusan Boyle has dropped yet another hint that she is gearing up to announce a brand new project with her latest post - after showing off a glamorous new look and teasing the start of a 'new era'. The singer, 65, has been hailed as a national treasure ever since she delivered a history-making performance of I Dreamed A Dream on Britain's Got Talent in 2009.
Read more »