Group Partition and Group Information based Multi-agent Consensus Systems

Consensus is a fundamental issue in distributed multi-agent systems, and utilizes the inter-agent relative state to update the state of each agent, so as to achieve global synchronization. Motivated by some scenarios such as privacy-preserving constraints, system measurement error, that the inter-agent relative states are not available, a novel strategy is required to achieve consensus. This paper proposes a consensus algorithm by group partition and state updating via relative group information replacing of the inter-agent relative state. In the scheme, a group of agents are divided into a couple of subgroups by certain probability. This paper proposes a sufficient and necessary condition to reach consensus in mean when subgroup number
$m = 2$
and also a sufficient condition when
$m = 2$
. Both numerical simulations and a smart-car hardware platform are provided to demonstrate the validity of our results.