Staircases are among the most challenging terrains a mobile robot can face. A multi-year field study found that robots designed for stair traversal fail at least 35 times more often on stairs than on level ground. The consequences can be significant. A robot that loses balance on a step accumulates momentum as it tumbles, threatening severe damage to itself, the building, and anyone in its path.
Fall prevention measures, such as path planning and balance control, can help the robot avoid noticeable hazards, but they cannot stop a person from accidentally walking into the robot from above. That residual risk is unavoidable, and it has kept operators from deploying heavy autonomous platforms in stairwell environments.
“This is why fall mitigation matters more than fall prevention alone,” said Professor Mohan Rajesh Elara, Engineering Product Development, Singapore University of Technology and Design. “Until the industry has a credible answer to that residual risk, operators will keep treating these platforms as a liability rather than a labor-saving tool.”
In their study, “A reinforcement-learning-based fall mitigation system for stair-traversing service robots,” Elara and his team developed a fall mitigation system for robots through the use of design, artificial intelligence (AI), and technology. They equipped a commercial-grade tracked robot with a three-jointed arm controlled by a policy trained entirely through reinforcement learning (RL) in simulation.
By surveying how falls happen on stairs, the team identified five fall modes: a straight backwards fall, two pivoting fall variants, and two sideways falls. They then explored the simplest type of articulated structure that could brace against all modes.
“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,” explained Prof 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.”
In each simulated episode, a force knocked the robot backwards or sideways, and the controller decided every fraction of a second how to move the arm joints. A proximal policy optimization algorithm then adjusted the controller's behavior based on the outcome by rewarding stable, upright finishes, while penalizing flips, falls off the stairs, or unnecessary arm movement.
Across five trained controllers, the RL system achieved an average success rate of 69.4 percent in arresting the robot’s fall and returning it to a stable position. In contrast, a hand-coded heuristic baseline succeeded only 38.6 percent of the time and often destabilized the robot further by flailing the arm. When the RL controller did catch a fall, it stabilized the robot in an average of 4.25 seconds — well within the 10-second window the team set as their target.
The most commercially significant result came from robustness testing. The controller, trained on a single robot and staircase geometry, was tested without retraining on platforms that were either 10 percent larger or smaller, and on staircases with altered step dimensions. On a larger robot, the best controller’s success rate climbed to 87 percent. On a smaller, less stable platform, it dropped but still functioned.
“The controller is not memorizing one geometry,” added Elara. “It is learning a recovery strategy that generalizes.” This means a single module could be reused across a family of robots sharing the same morphology, without retraining for each variant.
The researchers acknowledge that an average success rate of 69.4 percent falls short of what safety standards like the IEC 61508 demand for a standalone safety function. For the solution to reach deployment readiness, the team needs to raise the controller’s own performance, add compensating measures such as mechanical brakes and upstream fall-prevention layers, and meet explainability requirements through a surrogate model. A physical validation will first entail a simplified test rig, then integration onto a full platform, followed by testing across real staircases of varying geometries.
“Our goal is to make the system one credible layer of defense within a larger safety architecture. If we can demonstrate that it works reliably and is auditable, we move stair-traversing robots closer to being platforms that operators trust rather than fear,” said Elara.
Here is an exclusive Tech Briefs interview, edited for length and clarity, with Elara.
Tech Briefs: What was the biggest technical challenge you faced while developing this fall mitigation system?
Elara: The biggest challenge was identifying and addressing the wide range of scenarios that could lead to a robot falling while operating on staircases. These risks can be intrinsic, such as control errors, sensor uncertainty, or wheel slippage, and extrinsic, such as varying staircase geometries, surface conditions, obstacles, or unexpected interactions with people and the environment. Developing a system that can not only predict these potential fall events but also respond safely and recover robustly under real-world conditions was a significant research challenge.
Tech Briefs: Can you please explain in simple terms how it works?
Elara: The staircase-cleaning robot is a reconfigurable autonomous robot designed to clean one of the most physically demanding and ergonomically challenging cleaning tasks — staircases. Unlike conventional cleaning robots that operate only on flat floors, this robot can detect staircases, align itself with each step, reconfigure its body to climb stairs, and systematically clean the staircase while moving safely between steps. It achieves this using a combination of sensors, including LiDAR, distance sensors, contact sensors, and inertial measurement units (IMUs), which allow it to understand its environment and monitor its own movement.
The fall mitigation system acts as the robot's safety intelligence layer. As the robot climbs and cleans, it continuously evaluates both its internal state — such as stability, orientation, and motion — and external conditions such as staircase geometry, surface characteristics, obstacles, and environmental disturbances. AI-based models analyze these inputs in real time to identify situations that could potentially lead to a fall.
When a risk is detected, the system can proactively initiate corrective actions, such as adjusting the robot's posture, modifying its trajectory, reducing speed, or activating recovery behaviors. In essence, the robot is constantly asking itself, "Am I safe to continue?" and taking preventive action before a fall occurs. This capability is critical for enabling reliable deployment of staircase cleaning robots in real-world buildings where no two staircases are exactly the same.
Tech Briefs:Do you have plans for further research?
Elara: Yes. Our next phase of research focuses on scaling up testing and moving closer to real-world deployment. We are expanding the system's perception capabilities through multi-sensory sensing and AI-based prediction models that can better anticipate fall risks under diverse operating conditions. We are also investigating additional safety mechanisms, including mechanical safeguards and recovery strategies, that enable the robot to safely overcome challenging scenarios. The ultimate goal is to validate these technologies on full-scale staircase-cleaning robots operating across staircases with varying geometries and environmental conditions.
Tech Briefs:Do you have any advice for researchers aiming to bring their ideas to fruition?
Elara: Robotics is progressing at an unprecedented pace, but its true significance lies not only in what robots can do, but in how they can positively transform industries, workplaces, cities, and everyday lives. Researchers should stay grounded in real-world problems, engage closely with industry partners and end users, and focus on solutions that can create meaningful societal impact. Our work on staircase cleaning robots is one example of how robotics and physical AI can improve worker safety, address labor-intensive tasks, and enable the safe deployment of autonomous systems in challenging built environments.
Tech Briefs:Is there anything else you'd like to add?
Elara: Yes:
- Staircase cleaning remains one of the most ergonomically demanding and physically challenging cleaning tasks. Safe robotic automation has the potential to significantly improve worker safety and productivity.
- Fall mitigation is not just a robotics problem; it is a critical enabler for deploying autonomous systems in real-world environments where safety and reliability are paramount.
- The broader vision is to create Physical AI systems that can perceive, reason, adapt, and operate safely in complex built environments, enabling the next generation of intelligent service robots.
