Simulations of the lower bound on EDRb

The simulations regarding lower bound on EDRb are also implemented in AWGN and Rayleigh block fading channel. The lower point of each curve in Fig. 3.10 and 3.11

corresponds to the most energy efficient point which reveals that the increase of node density is helpful to let the network performance on energy efficiency achieving the

theoretical lower bound.

In Fig. 3.12, simulation results are given for different transmission distance. Here 400 nodes are deployed in the simulation area. This result indicates that the theoretical lower bound on the energy efficiency EDRb is valid for a Poisson network though its derivation is based on a linear network.

3.6

Summary

In this chapter, applications of the lower bound of energy-delay tradeoff are presented. A parameter optimization process is found to adjust the parameters including physical layer and protocol layers according for the applications with or without delay constraint. Meanwhile, the simulations in a 2-dimension Poisson network is provided to verify the theoretical results about the lower bounds of energy-delay tradeoff and energy efficiency. In simulations, two geography routing schemes are used, while this framework can be used for non-geography routing schemes as well because the transmission distance is an inherent property all routing schemes.

Cooperative communications

4

State of The Art

Channel fading was traditionally considered as a source of unreliability that has to be mitigated in wireless networks. However, information theory [32] reveals that chan- nel fluctuations can be rather beneficial if strong channel states are opportunistically exploited. To achieve the full capacity of such a system, a virtual Multiple Input Mul- tiple Output (MIMO) system [27] and an opportunistic communication scheme [90] are proposed.

4.1

Opportunistic communications

Compared with traditional point-to-point multi-hop routings, the basic idea of oppor- tunistic routing is that, at each hop, a set of next-hop relay candidates receiving a packet successfully compete for acting as relay as show in Fig. 4.1. For relay selec- tion, a priority is assigned to each relay candidate according to a specified metric, for example, the geographical closeness of the relay candidate to the destination [90]. As a consequence, exploiting the spatial diversity is exactly the purpose of opportunistic routing techniques.

The efficiency of opportunistic communications can be evaluated from different points of view. In this part, we keep our analysis approach in Chapter 2: the end to end reliability is a hard constraint while the end to end delay and the ene to end energy consumption are two interaction factors.

In classical opportunistic communications, the neighbors of a source in a certain 57

58 Opportunistic communications

S

D

Best relay

Best relay

Figure 4.1: Mechanism of opportunistic communications

area serve as the relay candidate. For example, the nodes in the sector with a angle ψ of S are the relay candidates of S as shown in Fig. 5.1. Some analytical models have already been proposed for this kind of opportunistic communications in [70,56]. The modeling framework in [70], which separates the opportunistic routing functionality into three components: routing, medium access and sleep discipline, is proposed to analyze the energy efficiency and latency performances of opportunistic routing in low traffic scenarios. This framework rests on the disc model [35] which relies on the definition of a radio power reception threshold and can not take the realistic channel features into account, such as fading or shadowing. In [56], the proposed model introduces shadowing and fading and defines the expected effective transmission distance (ETD). However, this model still relies on the definition of a reception threshold, now defined as a signal to noise (SNR) threshold, namely, the switched link model. As described in Chapter1, the disc model and the switched link model are not realistic and have severe weakness that are particularly relevant to opportunistic communications. In addition, the expected ETD model in [56] is a rough approximation of reality by the summation computation. Furthermore, none of the aforementioned models considers the impact of retransmissions caused by link unreliability. In chapter 5, we propose a significant improvement of the expected ETD model proposed in [56].

On the basis of the models previously described, several works provide a thorough analysis of the energy performance of opportunistic routing. In [90, 70], energy and latency performances of an opportunistic routing scheme called GeRaF are analyzed, and the effects of node density, traffic load and duty cycle are evaluated. The simu- lations in [71] show the impact of node density, radio channel quality and traffic rate on the energy consumption at each node, the average packet transmission delay and the goodput of opportunistic protocol. A 10% decrease in power and a 40% of delay reduction is exhibited in this work. Whereas these analyses are based on the previously described unrealistic disc link model. In addition, the energy efficiency of an oppor- tunistic routing called CAGIF [86] is studied in a fading channel, where the whole set of neighbor nodes try to receive the packet from the source node.

However, none of the aforementioned works consider the optimization of the trans- mission power and the number of relay candidates, which have influence on the energy performance. Consequently, these studies are insufficient to determine whether the

relative low performances of opportunistic routing are intrinsic to this kind of rout- ing or due to the specific protocol implementation (relay selection policy, fixed power choice, etc...). In Chapter 5, we will analyze the minimum energy consumption with opportunistic communications without delay constraint in classical opportunistic com- munication.

Several previous works on opportunistic routing, such as [90,70,71,86,85], provide the analyses of energy and latency performances. In [90,70], energy and latency perfor- mances of a routing scheme called GeRaF are analyzed, and the effects of node density, traffic load and duty cycle are evaluated. The simulations in [71] show the impacts of node density, radio channel quality and traffic rate on the energy consumption at each node, the average delay of packet and the goodput of opportunistic protocol. It is concluded that the benefit of opportunistic scheme is about 10% decrease in power and 40% reduction in delay. Whereas these analyses are based on an unrealistic disc link model [83, 35] which relies on the definition of a reception threshold level and is not well adapted to the research of opportunistic communications due to the neglect of propagation phenomena, e.g., fading and shadowing. Furthermore, the energy effi- ciency of the protocol CAGIF [86] is studied in a fading channel, where the whole set of neighbor nodes try to receive the packet from the source node.

However, in aforementioned studies, a fixed transmission power and the node density are considered, without providing any proof of optimality. Therefore, these studies are insufficient to determine whether the relative low performances of opportunistic routing are intrinsic to this kind of routing or due to the specific protocol (relay selection policy, fixed power choice, etc.). In Chapter 5, we propose to answer this question by minimizing the energy consumption of opportunistic relaying schemes with respect to the optimization of the node density and the transmission power.

In the mentioned works, all nodes in the communication area of a source node act as forwarding candidates. While in [85], instead of choosing the whole neighbors as relay candidates, an efficient selection mechanism of relay nodes is proposed in order to optimize energy efficiency. Simulations of [85] in a shadowing channel indicate that the energy efficiency is greatly improved.

Nevertheless, based on the relay selection mechanism, the optimization of param- eters is not considered and the delay performance of a system is not analyzed in the papers mentioned. In Chapter 6, these works are performed on the basis of a realistic unreliable link model.