Distributed strategies for minimum-latency cooperative retransmission in wireless networks

We consider cooperative retransmission strategies in wireless networks, where the retransmission of a failed frame is handled not by the original source but rather by common neighbors overhearing the transmission. The majority of existing literature in this space focuses on opportunistic mechanisms for choosing a single ¿best¿ neighbor, with a goal of minimizing the number of required (re-)transmissions. However, the coordination overhead of such mechanisms renders them unsuitable in general for scenarios involving delay-critical control or sensing applications, where the delivery latency, rather than number of retransmissions, is the dominant performance criterion. Accordingly, we study a distributed uncoordinated setting, where each neighbor that successfully overhears a frame decides independently whether to retransmit it in subsequent time slots, considering that multiple simultaneous such retransmissions will cause a collision. We employ a Bayesian approach to analyze the evolution of the system state view from the perspective of each cooperative neighbor, and derive a strategy of finding a sequence of retransmission probabilities for every neighbor in each time slot to minimize the expected delivery latency. We demonstrate for a wide variety of scenarios that this strategy achieves a significantly lower expected latency than either traditional retransmission or two-hop routing to the destination.