Fault-tolerant implementation of a distributed MLE-based time synchronization protocol for wireless sensor networks

This paper describes the implementation and evaluation of R⁴Syn protocol on MICAz platform and TinyOS operating system. The contribution is two folds. First, the implementation uses thorough maximum-likelihood estimators (MLE) in the joint offset/skew model, while all similar MLEbased estimators are merely evaluated with theoretical and numerical analysis thus far, and empirical solutions use simple computation estimators, such as offset-only models, or linear regression for skew estimation. Difficulties that has been encountered and overcome are reported in this paper. The second contribution is to consider fault-tolerance, an aspect that has been completely abstracted in previous works. The implementation assures correct behavior despite nodes failure or packet loss, as demonstrated by the experiments. Experimental results also demonstrate microsecond-level precision and long-term validity of the estimators in the joint skew/offset model.