Localization for mobile platforms, in indoor scenarios, represents a cornerstone achievement to effective develop service and field robots able to safely cooperate. This paper proposes a methodology to achieve such a result by applying a completely decentralized and distributed algorithm. The key idea of the solution developed is to enable a dynamic correction of the position estimate, computed by robots, through information, shared during random rendezvous. This objective is reached using a specific extension of the Extended Kalman Filter, called Interlaced Extended Kalman Filter, which allows exchanging the estimation performed by any single robot together with the corresponding uncertainties. The proposed unsupervised method provides a large flexibility: it facilitates the handling of heterogeneous proprioceptive and exteroceptive sensors, that can be merged taking into account both their accuracy and the system model one. The solution is particularly interesting for rescue scenario, since it is able to cope with irregular communication signals and loss of connectivity among robots team without requiring any synchronization.