When considering decentralized systems, complete knowledge of the state can typically not be assumed. For this reason, several strategies have been developed, in the literature, that let each node estimate the overall state of the system, based on local knowledge only.
This can be applied to several domains, such as multi-robot systems, sensor networks, etc.
We are considering the estimation problem in multi-robot systems, with the objective of having single robots able to estimate global properties about the system based on information acquired by neighboring robots. Such information include topological information about the interconnection graph, or the evolution of the state of other robots.
Such estimation can then be used for different purposes. For instance, it can be exploited for enforcing global properties (e.g. connectivity, bi-connectivity). Another application is fault identification and isolation (i.e. understanding if one robot is misbehaving).
Imperial College, London
(Dr. Francesca Boem)