This thesis investigates reliable multicast of bulk data in a resource-constrained mission- critical radio access network (RCMCRAN). Traditionally, bulk data is sent over-the-air using reliable unicast techniques such as the transport control protocol (TCP). This suffers from poor scalability because the same transmission must be repeated to each member of the destination group. This problem is addressed by reliable multicast protocols. The main challenge with implementing multicast into RCMCRANs is that reliable multicast protocols are extremely application-specific. This research utilises existing techniques from literature and innovative methods to devise a reliable multicast error-control code for bulk data dissemination in RCMCRANs.

An iterative bulk data dissemination solution is designed that utilises single-layer hybrid automatic repeat request (HARQ) at the application layer of the protocol stack. The protocol’s error-control coding logic is designed to iterate between data dissemination using the Raptor forward error correction (FEC) code and a novel multicast feedback scheme. The Raptor code is selected because the set of receiving stations can recover from different symbol losses using the same set of new encoded symbols. The novel multicast feedback scheme is designed to facilitate guaranteed delivery. The Raptor code and feedback scheme aspects and error control logic are selected and sized to interface with each other correctly and to protect against the unique error characteristics of the RCMCRAN environment.

Within this study, research is focused on the feedback component of the error-control code. A novel automatic repeat request (ARQ) technique is presented. It utilises feed-back messages that comprise of bundled negative acknowledgements (NACKs). Remote stations within the multicast group distribute their feedback transmission timers across a feedback round and use random-access techniques to gain channel access. The re- mote stations assist each other in conveying their unique feedback messages to a shared destination. Stations overhear other feedback messages on the channel and combine them with their own. The peer overhearing provides an alternate transmission path from the source to the common destination that can recover feedback messages that were initially discarded at the destination due to collision or random error.

To facilitate the combining of feedback messages, different bundling algorithms were analytically evaluated. We show that the space-efficient Bloom filter that has previously been used in database optimisation is not suitable for storing retransmission requests within an ARQ scheme. Their size benefits diminishes when hashing a non-trivial number of NACK block identifiers into them. Instead, a bit-vector structure is found to be better suited due to its superior size, complexity, and error characteristics when implemented within the ARQ scheme.

Results for the ARQ scheme’s performance are given using analytical models and simulation results over feedback sessions in RCMCRANs with at most 1,000 remote stations in the group. Simulation results are collected using a custom-built Python simulation environment. Simulation results show that the ARQ scheme achieves reliable delivery of feedback objects from a multicast group of remote stations to a single base station. The ARQ scheme does this whilst being a scalable solution and avoiding the common multicast problem of feedback implosion. The addition of overhearing to the ARQ scheme reduced the average message error rate experienced due to its collision recover capabilities. We conclude that the overhearing technique improved the reliability of many-to-one transfer of feedback objects from a multicast group over a RCMCRAN. This work provides new insight into the use of overhearing as a collision recovery mechanism in random-access.