Multi-Input Ensemble Control Breakthrough: Neshaei Moghaddam et al. Solve Open Problem

Broadly, ensemble control tackles a deceptively hard problem: how to steer a very large population of slightly different dynamical systems using the same input signal. Think of delivering a precise stimulus to a diverse population of neurons, or synchronizing a fleet of drones with heterogeneous dynamics. A key question is whether a given system's sparsity structure (which states connect to which inputs) allows it to be 'averaged controllable' — i.e., for any ensemble of parameter variations, you can drive the average state exactly where you want.

The single-input version of this structural problem was fully solved in prior work, but extending it to multiple inputs turned out to be tricky: new coupling effects emerge, and the elegant graph-based conditions from the single-input case didn't directly generalize. Neshaei Moghaddam, Chen, and Gharesifard now close that gap. They prove a sharp necessary and sufficient condition: a multi-input linear ensemble system is structurally averaged controllable if and only if it is accessible (roughly, every state dimension can be excited by some input) and its associated acyclic subgraph — the 'core' — contains a row-saturating matching. This is a purely combinatorial condition, meaning you can verify it with bipartite matching algorithms without solving any differential equations. The result turns an intractable infinite-dimensional control problem into a polynomial-time graph check, opening the door to more systematic design of multi-input ensemble controllers for applications like neurostimulation and quantum control.