The design of decentralized controllers coping with the typical constraints on the inter-robot sensing/communication capabilities represents a promising direction in multi-robot re- search thanks to the inherent scalability and fault tolerance of these approaches. In these cases, connectivity of the underlying interaction graph plays a fundamental role: it represents a necessary condition for allowing a group or robots achieving a common task by resorting to only local information. Goal of this paper is to present a novel decentralized strategy able to enforce connectivity maintenance for a group of robots in a flexible way, that is, by granting large freedom to the group internal configuration so as to allow establishment/deletion of interaction links at anytime as long as global connectivity is preserved. A peculiar feature of our approach is that we are able to embed into a unique connectivity preserving action a large number of constraints and requirements for the group: (i) presence of specific inter-robot sensing/communication models, (ii) group requirements such as formation control, and (iii) individual requirements such as collision avoidance. This is achieved by defining a suitable global potential function of the second smallest eigenvalue λ2 of the graph Laplacian, and by computing, in a decentralized way, a gradient-like controller built on top of this potential. Simulation results obtained with a group of quadorotor UAVs and UGVs, and experimental results obtained with four quadrotor UAVs, are finally presented to thoroughly illustrate the features of our approach on a concrete case study.