Stay informed and ahead with a curated selection of content delivered straight to your inbox, ranging from daily news and career advice to monthly illuminations on AI, sustainability, software, and more. Dive deep into the cutting-edge engineering revolution that is reshaping modern defense with Strength, Stealth, Speed: The Very Fast Future of Advanced Defense. Enjoy expert insights, exclusive content, and a closer look at engineering and innovation with minimal ads or opt for an ad-free experience. All Rights Reserved, IE Media, Inc.
Amidst advancements in robotics, a significant breakthrough has been made in enhancing the safety of autonomous systems, particularly robots maneuvering staircases. Addressing the prevalent issue of robots losing balance and tumbling down stairs, researchers at the Robotics and Automation Research (ROAR) Laboratory at the Singapore University of Technology and Design (SUTD) introduced a novel approach. Unlike conventional safety systems that focus solely on fall prevention, this new system emphasizes fall mitigation by leveraging reinforcement learning and a 3-joint stabilizing arm to aid robots in recovering from falls on stairs.
Despite progress in balance control and path planning, robots designed for staircase traversal are still susceptible to frequent failures. Studies have shown that robots encounter over 35 times more failures on stairs than on flat surfaces. The inherent risk of falls is exacerbated by the rapid build-up of momentum once a fall commences. Existing safety mechanisms, such as navigation algorithms and obstacle detection, fall short in addressing unforeseen factors that may lead to a fall. Professor Mohan Rajesh Elara, leading the ROAR Laboratory, stresses the importance of prioritizing fall mitigation alongside prevention to enhance the operational efficiency of robotic platforms.
The newly developed reinforcement learning-based system enables robots to dynamically respond during a fall, utilizing a mechanical arm for stabilization. By studying various fall modes and designing a versatile robotic arm capable of countering these motions, the research team has significantly improved fall recovery. Through simulation-based training, the system achieved a remarkable average success rate of 69.4% in preventing falls and restoring stability, surpassing traditional control methods.
While the system exhibits promising results, further refinement is necessary to meet stringent industrial safety standards before real-world deployment. Ongoing efforts include enhancing reliability, incorporating fail-safe measures, and developing models for certification. This study, featured in the journal Engineering, underscores SUTD’s commitment to advancing the safety of autonomous mobile robots in practical settings.
Neetika Walter, an accomplished journalist with a rich background in covering diverse topics, brings her expertise to elucidate complex subjects with clarity and nuance. When not immersed in the world of journalism, she finds solace in books, poetry, and the companionship of her dogs, embodying a passion for storytelling and contemporary culture.
About the Author
