BraiNCA: brain-inspired neural cellular automata and applications to morphogenesis and motor control

2026-04-02

AI Summary

Traditional neural cellular automata (NCAs) are constrained by regular grids and strictly local neighborhoods, limiting their ability to model the long-range connections and complex topologies observed in biological brains. This work proposes BraiNCA, which for the first time integrates graph attention mechanisms and explicit long-range connectivity into the NCA framework. While preserving the principle of local update rules, BraiNCA effectively models irregular, complex network structures. This approach overcomes the limitations of purely local interactions, substantially enhancing system robustness, learning efficiency, and sample utilization. Experimental results demonstrate that BraiNCA achieves significantly faster convergence than conventional NCAs in morphogenesis and locomotion control tasks, while also exhibiting superior resilience to structural damage.

Abstract

Most of the Neural Cellular Automata (NCAs) defined in the literature have a common theme: they are based on regular grids with a Moore neighborhood (one-hop neighbour). They do not take into account long-range connections and more complex topologies as we can find in the brain. In this paper, we introduce BraiNCA, a brain-inspired NCA with an attention layer, long-range connections and complex topology. BraiNCAs shows better results in terms of robustness and speed of learning on the two tasks compared to Vanilla NCAs establishing that incorporating attention-based message selection together with explicit long-range edges can yield more sample-efficient and damage-tolerant self-organization than purely local, grid-based update rules. These results support the hypothesis that, for tasks requiring distributed coordination over extended spatial and temporal scales, the choice of interaction topology and the ability to dynamically route information will impact the robustness and speed of learning of an NCA. More broadly, BraiNCA provides brain-inspired NCA formulation that preserves the decentralized local update principle while better reflecting non-local connectivity patterns, making it a promising substrate for studying collective computation under biologically-realistic network structure and evolving cognitive substrates.

Problem

• Neural Cellular Automata
• long-range connections
• complex topology
• distributed coordination
• brain-inspired

Innovation

• Neural Cellular Automata
• attention mechanism
• long-range connections
• complex topology
• brain-inspired computing