The proliferation of robots like Boston Dynamics' Spot has demonstrated the versatility of quadrupeds. These systems are useful for climbing stairs, crossing small obstacles, and navigating uneven terrain. But ladders still have big problems. Especially considering that ladders are always present in factories and other industrial environments where the system is installed.
ETH Zurich, which has led some of the most exciting quadruped robot research in recent years, has set out its future direction. As the school points out, past attempts to tackle ladders have primarily used bipedal humanoid robots and specialized ladders, but ultimately proved too slow and ineffective. It turns out.
Investigation revealed that the school was once again using the ANYMal robot from its spin-off ANYbotics. Here, the team equipped the quadruped with a special end effector that hooks onto the rung of a ladder. But the real secret sauce is reinforcement learning, which helps the system adapt to different ladder characteristics.
The researchers said, “This study expands the scope of applications of industrial quadruped robots beyond nominal terrain inspection to difficult infrastructure functions in the environment, and allows robots to perform complex skills.” “We highlight the synergies between the form of control and control policies.”
The school said the combined system had a 90% success rate in navigating ladder angles ranging from 70 degrees to 90 degrees. It also reports a 232x increase in climb speed compared to current “state-of-the-art” systems.
The system can adjust climb instances to correct in real-time if you misjudge a run or mistime a step.