Abstract

This deliverable provides comprehensive theoretical and simulation results on the proposed power optimization algorithms with four toy scenarios identified in the project. The proposed algorithms mainly assume that the statistical channel knowledge and/or location information are available at each node, which aims to be in line with RESCUE “links-on-the-fly” concept. Specifically, the outage probability based joint power allocation and relay position for lossy-forwarding relaying scheme is firstly investigated in toy scenario one, and then the work is extended to a symbol-level selective transmission scheme. For toy scenario two, compared with our previous work in D2.2.1, the outage probability based power allocation is extended to multi-relay case, meanwhile, the power allocation from rate distortion perspective is also investigated.

Furthermore, the outage probability based power allocation for toy scenario three is presented for the first time, and the orthogonal multiple access relay channel based power allocation is also illustrated for the case with more than two sources. Based on the provided results, the proposed algorithms exhibit improved performances by comparing with the conventional schemes, e.g., equal-power allocation.

Executive summary

Recall the Links-on-the-fly Technology for Robust, Efficient, and Smart Communication in Unpredictable Environments (RESCUE) concept that relays are allowed to decode-and-forward the received frames with specified level of errors, which aims to provide efficient and simple information transfer. In this case, the error propagation effects can be mitigated at destination with modified distributed turbo decoding by taking source-relay link correlation information into account. Alternatively, relays can also predict the positions of decoding errors in a frame and then null out them in order to mitigate the error propagation effects. Both of these strategies have a common assumption that channel feedback from reception node to transmission node is not allowed. Thus, the optimal power allocation cannot be based on the knowledge of instantaneous channel state information (CSI). However, we can still use statistical CSI and/or nodes’ location information obtained through long term observation and training. In this deliverable, a comprehensive review of the proposed power allocation algorithms with statistical channel knowledge and/or nodes’ location information for different identified toy scenarios is presented.

Firstly, the joint optimization of power allocation (PA) and relay position (RP) for lossy-forwarding relaying is proposed, where the objective is to minimize the system outage probability of toy scenario one (TS1). With the closed-form expression of the outage probability, we investigate adaptive PA with fixed RP, adaptive RP with fixed PA ratio, and joint optimization of PA and RP under total transmit power constraint. It is found that the proposed three algorithms outperform the equal PA, midpoint RP, and semi-adaptive optimization algorithms, respectively. Moreover, we also consider the optimal PA and RP for a symbol-level selective transmission at relay scheme. In this case, the optimal power allocation is to maximize the average received signal-to-noise (SNR) ratio at destination, where the SNR expression includes the derived probability of correctly predicted/forwarded symbols per frame at relay. It is shown that, within four presented relay locations, relay closed to destination provides the best average SNR performance, and its optimal power allocation happens when the relay is allocated with more power.

As investigated in D2.2.1, the power allocation in order to minimize the system outage for two relays based chief executive officer (CEO) problem provides better performance than the ones with equal power allocation. In this deliverable, we extend the work with three or more relays cases and propose a simple, yet effective power allocation scheme based on the Slepian-Wolf theorem. Moreover, we also assess the performance of the proposed power allocation for a practical joint decoding (JD) introduced in literature, and the improved performances in terms of average bit-error-rate (BER) can be observed. In addition, we also investigate the optimal power allocation for the lossy communication networks in toy scenario two (TS2). Specifically, we consider the power allocation from rate distortion perspective in order to achieve optimum distortion under total power constraints. The problem can be formulated as convex optimization framework and solved by using Karush-Kuhn-Tucker conditions.

In this deliverable, we introduce the optimal power allocation in toy scenario three (TS3) for the first time. Based on the derived outage upper bound presented in deliverable D1.2.2, we design the power allocation strategy to minimize the outage upper bound subject to the total transmission power constraints. It is shown that the proposed power allocation strategy can be asymptotically optimal at high SNR range. Similar as the ones for TS2, we also assess the performance of the proposed power allocation strategy for a practical JD scheme. An improved performance in terms of frame-error-rate (FER) is also observed.

Comparing with the work for toy scenario four (TS4) in D2.2.1, we generalize the power allocation problem of lossy forwarding based multiple access relay channel for two sources, single relay and common destination case to more than two sources case. Here, we propose a heuristic power allocation approach for fading channels with the average SNR of each link. The objective of the problem is to minimize the transmit power subject to rate constants, and the non-convex based power allocation problem can still be solved with successive convex approximation method. Numerical results show that the proposed method provides better performances than the conventional cyclic redundancy check based DF relaying in terms of power assumption and outage probability.

To sum up, the above mentioned power allocation algorithms for different toy scenarios are implemented mainly based on statistical channel knowledge and/or nodes’ location information, which can be obtained from long terms observation and training. This is consistent with RESCUE “links-on-the-fly” concept, where the signalling to guarantee reliable transmission for a specific link is not allowed in unpredictable environments.