End-to-end automation of behavioural experiments with a collaborative robot

Behavioural measurement is essential for neuroscience breakthroughs, yet automating tests with live animals proves tricky. Skilled handling, dexterity, gentleness and consistency are required during transfer and preparation of animals for assays. Such bespoke automation systems are rare, impeding widespread adoption. Introducing STELA, a robotic system that uses an off-the-shelf collaborative robot (cobot) to run a complete Drosophila behavior experiment workflow autonomously.

From fly culture vials to data collection, the cobot mimics human laboratory procedures in a familiar workspace. It handles fly culture vials, CO2 anesthesia, sex classification using vision, single fly manipulation, and loading and disassembly of a behaviour assay, all in a programmed loop. During a 24-hour day, STELA executed 16 experiments (1024 visual learning assays in flies), compared to a human who, in an 8-hour workday, accomplished 6 (384 assays).

Fly retention suitable for analysis remained nearly identical (93.85% for STELA vs. 83.33% for humans), but with 12 times better handling consistency by the robot. In biologically comparable behavior experiments, both robotic and manual handling yielded the same mean fly behavioural outcomes. However, the robot significantly reduced variability in baseline and learning behaviors.

STELA's success showcases the potential of standard robotic systems in flexibly automating one of the most challenging animal experiments. It offers a low-variance, high-throughput platform for scalable experimentation and a pathway to closed-loop, democratized robotic acceleration of experimental discovery.