We are excited to share a new preprint by Feng-Yu Yue, “Transversal Passivity: Dissipation Relative to Input and Output Manifolds.”
The paper introduces transversal passivity, a local dissipation framework for regulating input and output signals to prescribed manifolds. By measuring only errors transverse to those manifolds, the approach allows motion along them without penalty. The work develops conditions for transversal passivity, studies how the property behaves under feedback interconnections, and identifies geometric conditions that can obstruct convergence guarantees.
A federated learning case study with heterogeneous data illustrates how the framework can certify convergence even when the agents’ targets are not fully compatible with the interconnection.
Congratulations to Fengyu on this exciting new work! Read the preprint on arXiv.