Researchers at the Robotics Systems Lab at ETH Zürich have developed innovative techniques to overcome this challenge: The ladder climbing robot is equipped with a “hook end-effector ” and a “ reinforcement learning system ” and is able to climb stairs with inclines of up to 90°, achieving a success rate of 90%.
The advancement and development of the ladder climbing robot
This breakthrough and development is significant, as although “ legged robots ” have advantages over wheeled robots on uneven terrain, their inability to climb stairs limited their access to dangerous areas in industrial facilities. This limitation reduced productivity, and forced human workers to carry out inspections in risky areas. However, ETH Zürich ’s new solution improves safety, and increases the robot’s efficiency, allowing it to access previously inaccessible places.
The system integrates an “extension” equipped with a hook, which mimics the motion of a human grip when climbing a ladder. The robot uses reinforcement learning to hone its skill, including watching simulated videos of it facing various obstacles, such as unstable ladders or poorly positioned rungs. It was then tested in the real world, proving it could successfully climb in 90% of cases, even under unmodeled conditions, such as unforeseen disturbances.
Furthermore, the development team highlights that ” ANYmal achieved a climbing speed 232 times higher ” than previous bipedal robot solutions, which highlights the effectiveness of the integration between a new physical design of the robot and control policies based on learning. This technological advance improves the integration of robots in industrial infrastructures, allowing for safer and more efficient inspection in complex environments.
The quadruped robot and its robust climbing of stairs.
Source: Robotic Systems Lab: Legged Robotics at ETH Zürich
In the future, the researchers plan to continue refining the robot, integrating additional capabilities that would allow it to climb stairs in ” unstructured places ,” without the need for external motion capture equipment or other sensors.
Follow us on social networks and don’t miss any of our publications!
YouTube LinkedIn Facebook Instagram X
Photo: Robotics Systems Lab
Loading...
