Zhao's new algorithm solves matrix-weighted network controllability in polynomial time

Control theory gets a significant advance with Lanhao Zhao's new paper on strong structural controllability (SSC) of matrix-weighted multi-agent networks. Unlike scalar-weighted networks, matrix weights reflect higher-order dynamics where each node's state dimension exceeds one—common in robotic swarms, power grids, and distributed sensor systems. The paper establishes an upper bound on the strong structural controllable subspace (SSCS) using equitable partitions, then goes further by addressing two thorny realities: matrix singularity and asymmetric couplings. Zhao's solution decomposes the matrix-weighted network into layered scalar networks via a matrix space basis decomposition, effectively simplifying the problem while preserving the physical structure.

The core innovation is a layer-specific distance partition (LDP) that yields a tighter Squeeze Theorem, narrowing the mathematical boundaries of the controllable subspace by capturing layer-specific structural delays. To find the optimal basis that minimizes the gap between bounds, Zhao formulates an algebraic algorithm based on null-space projection. Using pattern matrices and generic rank, the paper rigorously proves that this optimal basis exists almost everywhere in parameter space—perfectly matching the definition of strong structural controllability. Finally, a polynomial-time automated discovery algorithm based on multi-layer Weisfeiler-Lehman (WL) color refinement eliminates the NP-hard combinatorial bottleneck of manually pre-defining targets. The paper also evaluates observability and derives invariant network attributes, supported by numerical examples with asymmetric weights and directed multi-layer topologies.