A team at the École Polytechnique Fédérale de Lausanne (EPFL) has unveiled a modular robotic hand that not only reproduces 33 human grasp types but extends them with dual‑sided grasping and autonomous crawling. Described in Nature Communications on 20 January, the device is a detachable “advanced hand” capable of handling objects outside the normal size envelope of robotic end‑effectors and of repositioning itself by crawling across surfaces to reach new targets.
The work addresses two stubborn problems in manipulation: versatility across diverse grasp geometries and mobility without relying on the whole robot to relocate. The hand’s dual‑sided grasping allows it to use more of its geometry to secure awkward or oversized items, while autonomous crawling gives it limited self‑repositioning ability that reduces dependence on arm or body movement and enables continuous sequential handling of multiple objects.
EPFL’s design is modular and appears aimed at practical deployment: the hand is removable, which simplifies maintenance and adaptation to different tasks. That modularity, combined with demonstrated multi‑target sequencing, suggests potential for integrated systems in factories, logistics hubs and service robots where hands must cope with variability and limited workspaces.
This advance should be seen against the broader progress in dexterous manipulation. Research groups and companies have long wrestled with reproducing the dexterity of a human hand; existing commercial and research hands reproduce many grasp types but usually trade off speed, robustness or mobility. EPFL’s contribution is not a finished product but a technical step that closes the gap between laboratory demonstrations and sustained, versatile operation in cluttered or unpredictable environments.
The immediate applications are tangible. In industrial automation the hand could reduce the need for bespoke end‑of‑arm tooling and simplify tasks that today require human operators or complex fixtures. In service robotics, an arm equipped with such a hand could better assist in homes or hospitals by handling objects of varying shapes and by repositioning itself in confined spaces. In field and planetary exploration, the ability to grasp oversized components and crawl into constrained areas could extend mission capabilities where full locomotion or human intervention is costly.
Practical hurdles remain. Robust, high‑speed tactile sensing, energy efficiency, reliable control across long operating periods and integration with perception stacks are necessary for real‑world deployment. The hand’s performance in adverse conditions, its durability under industrial duty cycles and the software maturity needed for unsupervised multi‑object tasks will determine commercial uptake. Moreover, as manipulation grows more capable, questions about workforce impacts, safety standards and dual‑use concerns will move from academic debate to policy forums.
The EPFL result is best judged as an enabling innovation rather than a market disruptor on its own. It demonstrates a promising architecture for more autonomous, adaptable manipulation but will need partnerships with manufacturers, systems integrators and software houses to translate lab performance into scalable products. For policymakers and industry leaders, the test to watch is not only whether the hand can replicate human grasps in controlled trials, but whether it can consistently perform in the messy, noisy environments that matter most to manufacturers, caregivers and explorers.