Differentially Private Average Consensus for Aperiodically Sampled-Data Heterogeneous Multi-Agent Systems: A Decomposition-Compensation Approach

This article addresses differentially private average consensus for heterogeneous multi-agent systems under aperiodic sampled-data control. A distributed hybrid dynamic controller based on a decomposition-compensation approach is proposed. Individual privacy is safeguarded by integrating non-decaying Laplacian noise into interaction information and decomposing the reference state into interactive and reserved substates, preventing direct exposure of true information to eavesdroppers. Leveraging the Robbins-Monro conditions, a time-varying step-size is designed to relax noise constraints, accelerate convergence, and achieve asymptotically unbiased output average consensus in mean-square sense. Finally, a numerical simulation and a physical experiments involving mobile robots and unmanned aerial vehicles validate the efficacy of the proposed scheme.