Abstract-In this paper, a relay selection policy is proposed that fully exploits the flexibility offered by the buffering ability of the relay nodes in order to maximize the achieved diversity gain. The suggested scheme incorporates the instantaneous strength of the wireless links as well as the status of the finite relay buffers and adapts the relay selection decision on the strongest available link by dynamically switching between relay reception and transmission. We show that the proposed relay selection scheme significantly outperforms conventional relay selection policies for all cases and ensures a diversity gain equal to two times the number of relays for large buffer sizes.
I. INTRODUCTIONWe focus on a simple decode-and-forward (DF) cooperative network where a source communicates with a destination through a set of half-duplex relay nodes with the objective of performance optimisation. For this fundamental problem, the max − min relay selection scheme [1], [2], where the selected relay holds the strongest end-to-end relaying path, achieves the optimal performance and ensures a full diversity equal to the number of the relays. However, this relay selection policy refers to a two-slot cooperative transmission where the selected relay receives the source's data during the first time slot and immediately in the next time slot forwards the data towards the destination. In order to overcome this limitation imposed by the max − min selection scheme, recently in [3] the authors introduced data buffers at the relay nodes that allow the selection of a different relay for reception and transmission in order to extract further diversity gains. In that work, the max − max relay selection scheme that selects the relay with the strongest source-relay channel and the strongest relaydestination channel for reception and transmission, respectively, has been investigated. The max − max scheme refers to applications without critical delay constraints and provides a significant coding gain in comparison to the conventional max − min selection scheme. However, despite the efficient use of the channel fading and the related performance benefits, the max − max relay selection policy requires a predefined schedule for the relaying transmission (e.g., the second slot is always allocated for relaying) and therefore the available diversity degrees are not fully exploited. It is worth noting that the use of buffers in order to boost the achieved performance (in terms of throughput) has been reported in the literature in different contexts ( [4]-[7] and references therein).