Data-driven consensus control for networked agents: an iterative learning control-motivated approach

This study considers an iterative learning control approach to achieve accurate coordination performances of the output data sequences for multiple plants that are involved in a networked environment. To realise such a desirable control objective, an update process of the input data sequence is needed to refine its output performance iteratively for each plant, which uses the local or nearest neighbour knowledge. The nominal multi-agent systems are employed as the plants' description, for which input-output data-driven consensus problems are addressed in a hybrid networked environment given by signed directed graphs with both cooperative and antagonistic interactions. It is proved that the output data can be guaranteed to achieve bipartite consensus or remain stable for the multi-agent networks under structurally balanced or structurally unbalanced signed graphs. Moreover, the convergence conditions are derived, which need less knowledge of the agents' plant, and the proposed consensus results can be developed to take into account the plant uncertainties and noises. Simulation tests are performed to verify the effectiveness of the learning approach in refining high input-output data-driven consensus performances of networked agents.