Robotic arms are generally affixed to floors or structures for stability due to their strength and speed, thereby limiting their range of motion. Researchers at EPFL’s Learning Algorithms and Systems Laboratory (LASA) aimed to expand the capabilities of robotic hands by developing a bimodal hand capable of detaching and operating independently from the robotic arm.

The newly developed hand possesses the ability to detach and reattach to a robotic arm as necessary. Typically, robotic hands are designed primarily for grasping objects. However, in an effort to create a hand that can both grip and move independently, similar to the character “Thing” from the Addams Family, researchers employed a genetic algorithm. This algorithm uses principles of natural selection and evolution to refine designs, supported by the MuJoCo physics simulator to evaluate the feasibility of various iterations.

Through these simulations, the researchers identified the optimal configuration: a hand with five articulating fingers, arranged similarly to a human hand. This robotic hand includes a magnetic connector at the wrist, enabling automatic attachment and detachment from the robotic arm.

The hand’s fingers can bend in both directions, a feature that facilitates independent locomotion and dual-object grasping when attached to an arm. This design enhances the hand’s functionality when mounted on an arm, allowing it to lift multiple objects simultaneously without needing to reposition unused fingers.

The innovative robotic hand is significantly smaller than other robots such as Boston Dynamics’ Spot, which uses four legs for locomotion. Although Spot has been upgraded with a robotic arm and grasper, an independently operating articulated hand could enable it to navigate and analyze areas that are otherwise inaccessible.

Source link