What is it about?
Modular robots are built from repeated building blocks. In this work we add a new ability to that system: “bones” whose length can evolve. We use large-scale simulations to see whether evolution actually takes advantage of this extra freedom and how it shapes both the body (morphology) and the behavior (controller). We compare different ways of setting goals (e.g., targeting certain body features while also encouraging useful behaviors). To check that our results are not just a simulation artifact, we select one high-performing, interesting robot and build its physical twin. By comparing how the simulated and real robots move, we learn about the reality gap (the differences between simulation and the real world).
Featured Image
Photo by Possessed Photography on Unsplash
Why is it important?
Letting evolution change size (not only shape) opens up new designs and can produce robots that move better and are easier to build. Our comparisons show when and how variable-length bones are exploited, offering practical guidance on how to set fitness goals when evolving bodies and brains together. By constructing and testing a real robot, we also provide evidence about sim-to-real transfer, which is essential for taking evolved designs out of the simulator and into real applications.
Perspectives
What excited us here was seeing evolution reliably use adjustable bone length to achieve desired structures and behaviors. Building the physical twin was key: it confirmed the core ideas and showed where motion in the real world diverges, helping us understand the reality gap.
Maarten Stork
Read the Original
This page is a summary of: Size Matters: Variable-Length Bones in Evolvable Modular Robots, July 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3712255.3726713.
You can read the full text:
Contributors
The following have contributed to this page
