Low-Complexity Buffer-Aided Link Selection With Outdated CSI and Feedback Errors

Abstract

Buffer-aided relays can improve the diversity of multi-hop networks, however, they increase packet delays. Thus, various delay-aware protocols have been developed, but without considering the transmission diversity. Moreover, most works adopt ideal assumptions, such as symmetric links, perfect Channel State Information (CSI) and error-free feedback channels. So, we propose a Low-Complexity (LoCo) link selection algorithm, herein called LoCo − Link. The proposed algorithm may experience delays during CSI updates, and hence, by using outdated CSI its performance may deteriorate. To alleviate this issue, we next propose a distributed version of LoCo − Link (d − LoCo − Link) dealing with outdated CSI. In both algorithms, the source performs broadcasting towards multiple relays; when the packets are transmitted by a relay to the destination, they are discarded from all other relays. This coordination relies on feedback channels. For non error-free feedback channels, we propose a scheme in which the relays listen to the transmission of the best relay and drop duplicate packets. Results show that LoCo − Link surpasses other algorithms, by decreasing the delay in asymmetric networks. Moreover, d − LoCo − Link avoids diversity losses due to outdated CSI, while the effect of non error-free feedback channels is mitigated by taking advantage of the inter-relay channels.

Hybrid NOMA/OMA With Buffer-Aided Relay Selection in Cooperative Networks

“…The case of equal rate can be mapped to two cellular users requiring the same service from a base station, while the case of a high-rate user and a low-rate user might correspond to a cellular user and an IoT device sharing the same wireless channel. The considered OMA schemes include the LoCo-Link algorithm of [32], the {R→D} prioritization algorithm of [24] and the delay-aware algorithm of [26]. Also, apart from the two proposed algorithms, NOMA versions of the algorithms of [24] and [26] are evaluated, as well as the delayaware NOMA algorithm of [47].…”

“…Hence, we can treat these data as one packet in the network. As a result, the theoretical analysis of this work is similar, mutatis mutandis, to those of [31], [32], and can be cast as a discrete time Markov chain (DTMC) in which each state represents a possible state of the buffers. The main difference in the BA-NOMA algorithm is that the values of the transition probabilities of the DTMC differ from the source to the relays, since the source transmits packets for two destinations, and the same holds for the transition probabilities from the relays to the destinations, as packets are transmitted only when both links are not in outage.…”

“…By using broadcasting, more packets are available at the relays' buffers and lower delay is achieved. Then, the work in [32] proposed low-complexity (LoCo) link selection, focusing on practical challenges of BA-ORS, including asymmetric links, outdated CSI and feedback errors. LoCo-Link exploited broadcasting and {R→D} prioritization to improve the performance.…”

“…When the average channel SNR, denoted herein by φ, goes to infinity (i.e., φ → ∞), the probability of a {S→R} or a {R→D} link being in outage goes to zero, provided the buffer is not full or empty, respectively. The prioritization of the {R→D} links at high SNR makes the scheme deterministic, as it is the case of the LoCo − Link scheme in [32]. The main reason is the following: since {R→D} links are prioritized for transmission, there will be a time, say t, at which all buffers will eventually become empty.…”

“…Then, for SNR values between 10 dB and 15 dB, OMA outperforms NOMA, as the interference between the users' signals cannot guarantee that both signals are successfully decoded at each destination. Nonetheless, when increased transmit SNR is available, due to the robust selection of α, NOMA offers improved performance compared to OMA, as the users' power allocation is based on channel Outage probability BA-NOMA NOMA [47] NOMA [24] OMA [24] OMA [26] OMA [32] Average sum-rate [bps/Hz] BA-NOMA BA-NOMA/OMA NOMA [47] NOMA [24] OMA [24] OMA [26] OMA [32] asymmetries and the required rate. Again, it can be seen that no differentiation is observed within each multiple-access family, due to {R→D} prioritization.…”