Research
Researchers at EPFL and New York University have introduced ScaFi (Scalable Fish), a fish-inspired robot designed so that the same underlying design can be built at different sizes without starting over. The work, published in npj Robotics, addresses a persistent limitation in bio-inspired underwater robotics: most fish-like robots are built for a single size and a single job.
Propeller-driven underwater vehicles remain the standard for aquatic monitoring, but their spinning blades can snag on vegetation, disturb sediment, and frighten the wildlife they are meant to observe. That makes them a poor fit for shallow creeks, dense weeds, or close approaches to fish. Roboticists have therefore spent years developing machines that swim by bending their bodies instead. The catch, according to the team, is that scaling these designs up or down has typically meant rebuilding and retesting from scratch.
Right now, if you want to monitor a creek and then monitor a lake, you basically need two different robots, built and tested from the ground up.
ScaFi is modeled on fish such as cod and mackerel, which swim by concentrating most of their body bending toward the tail end. In nature, this swimming style spans an unusually wide range of body sizes, making it a natural template for a scalable robot. The project was led by Nana Obayashi, first author on the study, who is now an assistant professor of mechanical and aerospace engineering at the NYU Center for Robotics and Embodied Intelligence and conducted the work while completing her PhD in the Computational Robot Design Fabrication Lab.
Obayashi framed the motivation in practical terms, noting that the environments researchers care about do not come in one size, and arguing that the tools should not either. The study appears in npj Robotics, with the robot credited to the 2026 CREATE Lab at EPFL.
By demonstrating a fish-inspired platform intended to hold its swimming ability across small, medium, and large builds, the team points toward a future in which aquatic monitoring hardware can be adapted to a creek or a lake without a ground-up redesign. The source does not detail the specific scaling limits, performance metrics, or field trials of the platform.
Source: Robohub