Robots self assemble shapes using neural cellular automata and local rules
Distributed Swarm Morphogenesis with Neural Cellular Automata
RoboticsMultiagent Systems
Summary
Making groups of robots come together to form shapes is hard because each robot only knows about its nearby neighbors. The authors created a way for robots to discover where to attach next by sharing simple signals from a neural cellular automata method. They also improved how the robots learn to build shapes gradually, avoiding mistakes. Their tests with flying robots showed these ideas can work for building different shapes without needing a fixed plan ahead of time.
What this means in practice
- •For robotics engineers: Coordinate large groups of robots to autonomously assemble complex shapes using decentralized signals without prior site planning.
- •For drone operators: Use relative local observations and control to organize quadrotor swarms into structured formations safely and flexibly.
Authors
Ray Verma, Eliseo Ferrante, Anthony Tzes
Abstract
Swarm morphogenesis requires robots to assemble desired shapes through local interactions. We present a modular framework integrating neural cellular automata (NCA) with existing site selection and collision-aware control. Within this framework, we introduce a decentralized attachment-site discovery mechanism. Robots already in the formation update and share local NCA states to identify new attachment sites, enabling progressive growth without a predefined list of sites during deployment. To bridge the mismatch between conventional NCA growth and gradual physical assembly, we additionally propose offline sparse-retention fine-tuning that simulates partial attachment, together with a false-positive loss that penalizes invalid site activations. Simulations demonstrate formation of diverse lattice morphologies, scaling to several hundred robots, and substantial improvements in morphology accuracy from the proposed fine-tuning. Experiments with Crazyflie quadrotors validate the integrated execution pipeline using robot-relative observations reconstructed from external tracking.