The inherent properties of Wireless Sensor Networks (WSN) disqualify most classic methods targeting timeliness guarantees. Assumptions of such methods as well as a restrictive notion of timeliness borrowed from classic real-time systems clash with the indeterminism of realistic scenarios.
In this paper, we introduce a generalized notion of timeliness which allows to provide meaningful performance metrics under unreliable conditions, common in WSN. We present a probabilistic metric to capture the level of confidence for the timeliness performance without restricting its applicability. It consists of the estimation of the end-to-end delay distribution function by using current local state information of intermediate hops, which requires low memory and computational resources.
This metric represents a hook to adaptive QoS as it is constantly updated at run-time and reflects the actual network status. Extensive simulation results underline the validity of the method and its applicability.
