Time slotted channel hopping policies for industrial wireless networks

Future Industrial Internet of Things (IIoT) systems will require predictable, real-time, and reliable wireless solutions to connect the embedded devices used in process control industries. To meet these requirements, existing solutions commonly use Time Slotted Channel Hopping schedules that service a single link per slot and channel pair. To tolerate link variability, a common feature of harsh industrial environments, these schedules must incorporate retransmissions. However, because schedules can allow only one retransmission per slot and channel pair, a significant number of slots must be reserved for each link to ensure all possible failure scenarios are addressed. This is because retransmissions scheduled in isolation cannot be re-purposed in real-time in response to link variability. Since failures typically do not occur uniformly across a network in practice, existing industrial wireless protocols tend to incur much more latency and capacity costs than necessary to ensure packets are reliably delivered to their destinations. To overcome this limitation of existing industrial wireless solutions, this thesis works to develop and empirically evaluate Time Slotted Channel Hopping policies that can adapt in real-time to link dynamics.

This thesis makes the following contributions. (1) We propose a Flow-Centric Policy (FCP) that allocates retransmissions per flow in contrast to allocating retransmissions per link. At run-time, FCP will dynamically redistribute the allocated retransmissions for a flow to its links as needed depending on the transmission successes and failures observed. (2) We propose a method for configuring the number of retransmissions for FCP to achieve a user-specified end-to-end reliability and a scheduling framework to determine when retransmissions may be allocated and used so as to ensure all flows meet their latency requirements. (3) We propose a Star Policy (Starp) that allocates retransmissions to a select set of nodes within the network. At run-time, each of these nodes will act as a coordinator of a star topology by initiating and allocating retransmissions to its links as needed. (4) We propose a Threshold Link Reliability (TLR) model that can be used to guarantee a user specified end-to-end reliability for Starp policies even as links vary overtime arbitrarily. Moreover, we propose a mechanism by which to synthesis Starp policies using the TLR model such that all flows meet their reliability and latency requirements. (5) Finally, we perform extensive simulations and empirical experiments to evaluate both FCP and Starp and demonstrate that both can significantly improve network latency and capacity without loss of reliability.