Impact of Network Dynamics on Tardiness of Data in Sensor Networks

Impact of random delays and losses in sensor networks manifests in the form of tardiness of data used for processing at the sink nodes. The age of data used by the end application can impact the accuracy of the end results, and may produce detrimental consequences for many real-time sensing applications. This paper uses a tardiness measure for quantitatively capturing the lateness of the data due to network dynamics, and presents an analytical model relating the network delay, network packet loss rate, packet reordering, and sampling rate to the tardiness. We extend this model to provide aggregate weighted tardiness of data at fusion nodes. The tardiness model is validated using simulation results. We also investigate the tradeoffs between energy consumption and tardiness when desired tardiness is achieved by adjusting sampling rate and transmission power of the sensors nodes. Other potential applications of proposed tardiness measure includes determination of the active/sleep period of MAC layer to meet application goals, and comparison of routing protocols based on their impact on tardiness, and hence the application in real-time sensor networks.