Robustness in multi-robot systems

Abstract

Decentralized control strategies for multi-robot systems typically assume the presence of a connected communication graph among the robots. This allows to assuming the possibility of exchange of information, so that robots can complete common tasks, and achieve common objectives.

While strategies exist for ensuring connectivity preservation as the robots move, typically the presence of failures is not considered. However, when dealing with real robotic systems, failures can not be neglected. As is well known from the literature on complex networks, failures of the most central nodes can lead to fragmentation.

Our main idea is to propose methods for guaranteeing that, even in the presence of failure of a certain number of central nodes, the network remains connected.

We are considering two different approaches:

• An analytic approach, in which we consider the concept of bi-connectivity. Using eigenvalue conditions, we guarantee the preservation of the bi-connectivity property, i.e. the presence of two alternative paths among each pair of robots.
• A heuristic approach, in which each node estimates, based on local knowledge, its probability of being in a vulnerable configuration (i.e. about to lose connectivity), and subsequent mitigation policies.

People