EPFL · Wetenschap

Robotic Fish Scales Across Sizes for Aquatic Study

Researchers from EPFL and New York University have developed a scalable robotic fish, named ScaFi, capable of maintaining its swimming ability across various sizes. This innovation aims to provide a versatile tool for studying diverse aquatic environments, from shallow creeks to open waters, overcoming limitations of traditional propeller-driven underwater vehicles.

Traditional propeller-powered underwater vehicles are limited by their mechanics, which can snag on vegetation, stir up sediment, or startle wildlife, making them unsuitable for sensitive or complex aquatic environments. Roboticists have explored fish-like robots that swim by bending their bodies, but these are typically designed for a single size and purpose, requiring extensive redesign for different scales.

The ScaFi robot is modeled on fish like cod and mackerel, which swim by concentrating body bending towards the tail. This natural swimming style spans a wide range of body sizes, inspiring the robot's design. The engineering team found that by proportionally thickening the fiberglass rods in the tail as the robot scales up, while keeping the underlying motor mechanism and crossed-tendon system consistent, they could preserve similar tail-bending behavior.

This scalable approach reduces the engineering effort required to build robots for different environments and allows for systematic study of how swimming performance changes with scale. The team built three ScaFi robots (0.6, 1.1, and 2.9 meters long) and tested their swimming. The smallest robot produced water patterns similar to real fish, and all sizes showed comparable swimming motion when adjusted for body size, indicating the scaling preserved the fish-like gait.

Field tests included deploying the medium robot in a Swiss stream and the largest on Lake Geneva, with the smallest tested in creeks as shallow as 15-30 centimeters. While the swimming motion scaled well, energy efficiency did not. The largest robot was less efficient, suggesting drag and inertia may be factors. Agility also varied with size; the smallest robot was most agile but recovered slowest from disturbances, while larger robots were less nimble but more stable. The researchers suggest this scaling approach could be applied to other compliant robots, though scaling energy efficiency remains an open question.

AI-samenvatting op basis van de bron.