Topology Inference on Partially Observable Mobile Robotic Networks under Formation Control

Abstract

Mobile robotic networks (MRNs) have received considerable attention in recent years. Among their wide applications, the interaction topology plays a critical role in achieving efficient cooperation. We demonstrate that the observable interaction process of MRNs under formation control will reveal the interaction topology, presenting increasingly viable threats to the security of MRNs. To practice, we focus on a general situation where the external observer has only partial observation over the global MRN and aims to infer the interaction topology. This is a very challenging problem because the observable state evolution is a coupled consequence of the underlying interaction topology, the unknown leader dynamics, and the unobservable part. To address this issue, we novelly identify the steady pattern of the MRN and filter the impacts of the observable parts by shrinking the inference range. By constructing the informative dataset reflecting the interaction process, the local interaction topology is approximated with high accuracy. Simulations illustrate the feasibility and effectiveness of the proposed method.