Prescribed-Time Average Consensus Through Data-Driven Leader Motion

In this paper, a novel approach for finite-time average consensus within a prescribed time is introduced and explored for a multi-agent system running ordinary nearest-neighbor protocols. Differently from the available approaches of finite-time consensus, it exploits the special motion of one agent properly tuned, through an appropriate algorithm based on past values of its own local data. The leader's motion influence input is built upon a set of internal variables, and it is designed to achieve a finite-time transient behavior while keeping the initial consensus value unchanged. A preliminary procedure is set to retrieve the algorithm's parameters from data collected in an initial experimental stage, so that the overall algorithm can be run in a fully data-driven fashion. Simulation results put in evidence the effectiveness of the approach, together with some features and drawbacks of the proposed algorithm, so they draw some important insights regarding its use and future research lines.