Swift progress for robots over complex terrain

TL;DR

This paper demonstrates a four-legged robot that efficiently and rapidly navigates complex terrain like sand, outperforming humans in speed with low energy consumption by integrating data-driven learning with simple models.

Contribution

It introduces a novel approach combining data-driven learning and simple models to enable fast, energy-efficient robot locomotion over complex terrain.

Findings

Robot runs faster than humans on sand

Low energy consumption achieved

Few failures in terrain navigation

Abstract

A four-legged robot has learned to run on sand at faster pace than humans jog on solid ground. With low energy use and few failures, this rapid robot shows the value of combining data-driven learning with accurate yet simple models.