Original title: Fingers as Legs: Learning Self-Supported Locomotion and Manipulation with an Anthropomorphic Hand
Amirhossein Kazemipour, Hehui Zheng, Robert Katzschmann · September 2026
Most walking robots rely on separate wheels or legs to travel, plus arms or claws to move objects. Combining these parts makes robots bulky. Researchers wanted to see if a human-shaped robotic hand could carry its own weight and walk using only its fingers, while still performing handy tasks. The robot carries its own power and computer. The team trained it using reinforcement learning, a trial-and-error method, inside a computer simulation matched to real-world measurements. Their system accounted for uneven finger lengths. It learned to crawl, steer, and recover from falls on real hardware. While holding itself up, the walking hand pressed keys on a keyboard in a row without seeing them, and it pushed an object to targets using an overhead camera.
Imagine using your hands to crawl across the floor while also trying to pick up objects. Normally you would use your legs to move and your hands to grab, but here the same fingers must do both. In this paper, the robotic hand uses its fingers to both walk and manipulate, so it has to learn how to balance and move without separate legs.
This work shows that a robotic hand can be a complete mobile manipulator without extra legs or wheels. It simplifies robot design for tasks like search and rescue or household assistance, where a compact, self-contained device can crawl, steer, and interact with objects. The hand reuses its fingers for both movement and manipulation, reducing mechanical complexity.
The abstract doesn't explicitly state limitations, but it notes that the policies are task-specific and object pushing requires overhead visual feedback. The hand's locomotion is limited to crawling and steering, not walking upright. The approach relies on simulation training and may not generalize to all environments.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Fingers as Legs: Learning Self-Supported Locomotion and Manipulation with an Anthropomorphic Hand — Amirhossein Kazemipour, Hehui Zheng, Robert Katzschmann
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.