The robotic hand can detach from the arm, move towards objects, and pick them up.
The video was presented alongside a research article bearing the same title at the IEEE International Conference on Robotics and Automation, taking place in Rotterdam this week.
The recent video “Beyond Manual Dexterity” has captured attention at the IEEE International Conference on Robotics and Automation in Rotterdam. From its opening moments, the video showcases an extremely skillful robotic grip that leaves a striking impression. However, as it progresses, the narrative delves into uncharted territories. This research, developed by EPFL in Switzerland, aims to expand the capabilities of robotic hands to allow them to grip a greater variety of objects.
Researchers claim that their deep learning model has significantly improved manipulation techniques for multi-finger grips. However, they acknowledge that the topic of grip guided by contact information in cluttered environments remains a poorly explored field.
For decades, robotics specialists have drawn inspiration from the natural world, leading to the creation of quadrupedal and humanoid robots that flood the current market. While biology provides endless ideas, it can also constrain designers’ creativity. This raises the question: why should a robotic hand be restricted by the limitations of a biological model? The answer seems to depend on imagination.
The presented research challenges these limitations and demonstrates how systems can break these ties. However, the results can be unsettling. The creation of a hand inspired by the human one, capable of flexing backward to pick up various objects and even detaching to move autonomously, generates a disturbing effect, placing it in that uncomfortable zone known as the "uncanny valley." If it didn’t resemble a human hand so closely, it would likely evoke less discomfort.
Navigating this new world requires overcoming that visceral reaction. Beyond this, a promising future for robotic grips is in sight, although it remains to be seen if this approach can be effectively scaled. The research has sparked significant interest in the field, with researcher Xiao Gao playfully claiming that their team is probably the first to introduce this idea into robotics. Thus, biologically inspired robots are bid farewell, and "horror robots" are welcomed.
