Estimation and analysis of communication service time in a real-time wireless industrial network

Wireless communication systems for industrial applications should cope with the demanding performance figures of real-time wired networks, providing high reliability and satisfying tight timeliness constraints. The implementation of a scheduling framework may represent a relevant method to deploy real-time behavior in distributed measurement, process control and supervision networks. A schedulability analysis, based on task parameters, determines if the network will be able to provide the guarantees required by the specified application. Nonetheless, this strategy and the related error recovery mechanism are often implemented at a high level protocol layer, without carefully considering the behavior of physical/data-link layers of the network. An effective characterization of the service time for packet delivery, as seen by the data link-layer, is instead required for a fine tuning of the scheduling algorithms. We analyzed the case of an IEEE 802.11g based network, where a central scheduler manages the traffic. We derived a model for the transmission times, in particular accounting for retransmissions due to channel impairment and interference. An improved model for the service time is then proposed, based on a cross-layer analysis. This allows to reshape the statistical knowledge of the metric, hence improving the network performance and admission test outcomes. The provided case-study enables the reader to understand the effectiveness of the proposed method, which provide a reliable soft real-time data acquisition system.