Network Topologies 3
3.1.1 Energy Efficiency
As described in Section 2.1.3, energy consumption is an important issue in WBANs.
Equipping the sensors with replaceable or rechargeable batteries is not always possible, as this reduces the comfort of the person wearing them. Further, in Section 2.3.1 we have seen that the human body is a very lossy medium and the electromagnetic waves are attenuated considerably. This also induces tissue heat-ing [3, 4] which can arise when too much power is transmitted near the human body.
The energy efficiency can be measured by several performance indicators. In a WBAN, it is important that the network functions as long as possible. Further, all the nodes are equally important as no redundant nodes can be used (due to cost reasons and usability) and all the sensors generate and transport medical data.
Therefore, we propose to use the network lifetime as primary metric, which we define as the time for the first node to die, i.e. to run out of energy. In order to have a high network lifetime, the most consuming node should be made more energy efficient. To that end, we define the maximum network energy as the energy consumed by the most consuming node(s) of the network. This metric is inversely proportional to the network lifetime. When the maximum network energy is high, the lifetime will be low and vice versa. The network lifetime metric forces us to consider all nodes to be equally important.
By decreasing the distance between transmitter and receiver, the energy con-sumption of a node can be lowered. In Figure 3.1 an example is shown for the energy consumption of the Nordic nRF2401 radio with the parameter of Table 2.4.
The distance is calculated with (2.4) and a path loss η of 5.9. It is clear that by putting the nodes closer to each other, the energy consumption will drop. How-ever, we can not keep on repeating this process as at a certain distance the energy gain will be very small, i.e. when the distance is less than 30 centimeter. Indeed, the static energy of the transceiver will consume more energy than the transmitter.
The transmission distance between two nodes can be lowered by using intermedi-ate nodes as relay devices. In order to know whether or not the use of relay devices is more energy efficient, the static energy needed for forwarding the data needs to be considered.
A study for a line topology in sensor networks has been done in [5]. They used a similar radio model and found that multi hop communication is better when
CHAPTER3 51
Figure 3.1: An example of reducing energy consumption in a WBAN. By decreasing the distance that needs to be covered, the output energy can be set lower and the overall energy consumption will drop. The distance is calculated with (2.4) and the path loss η = 5.9.
the distance between the sender and receiver exceeds the characteristic distance dchar[6]. Using the notations from this thesis, dcharis defined as
dchar = η
sET Xelec+ ERXelec
Eamp· (η − 1) (3.1)
For the parameters of Table 2.5 for the Nordic chip, we see that dchar is 43 cm for a η = 5.9. This distance is rather small compared to the size of the body, so it is expected that a multi hop architecture will be the most energy efficient. The optimal number of hops can be then be calculated as follows when the distance between the sender and receiver is D:
Kopt =
D
dchar
(3.2) In a WBAN, the maximum distance between two nodes will never by higher than 2 meter. Thus, the optimal number of hops will never be higher than 5 with dchar = 43 cm. This definition for dchar however only considers a line topology and does not take into account the effects of aggregation.
In this work, the energy efficiency will be studied in several stages. First, we investigate the effects of a multi hop approach analytically for two different topologies: a line and a tree topology. Then the most optimal network topology is determined using linear programming. Next, relay devices are added to the net-work and the ILP-formulation is extended. In a last analysis, cooperation between relay devices and normal devices is investigated.
52 NETWORKTOPOLOGIES FORWBANS
3.1.2 Aggregation
While studying the energy efficiency, the effect of in-network aggregation will also be taken into account. Aggregation results in reducing the size of data that is forwarded by an intermediate node to its parent. This is an important topic for wireless networks in general. Data or packets are aggregated in order to reduce the number of packet transmissions. In wireless sensor networks, the measured data is highly correlated and aggregation is used to merge information. This merg-ing function can be a very simple mathematical function, such as max, min or average, or it can be a very complex algorithm. This data merging is not usable in a WBAN as each sensor will measure a specific physiological parameter. Fa-solo et al. give an overview of the aggregation techniques in WSNs [7]. Cluster based data-aggregation uses cluster heads which collect the data from surrounding neighbors. The cluster head performs local aggregation and sends the digest to the sink. A typical example is LEACH [8]. A second approach uses aggregation trees where data is aggregated while it is being routed over the aggregation path, e.g.
Directed Diffusion [9].
In this work, we will consider aggregation where the payload from different packets is fused and a new common header is added. Each data packet has a header that contains the address of the source, the destination, error correction codes etc. When the node forwards the packet, this header can be omitted when the data is sent together with the data of the node itself as this data already contains a header. The length of the packet when it is forwarded is thus reduced by a factor φ. This is referred to as the aggregation factor. For example, if a node i receives a packet of k bits from node j and aggregates it with its own data (l bits), the new packet size of node i is l + φk bits. When φ = 1, no aggregation takes place (e.g.
when the packet is sent separately). This type of aggregation is also referred to as concatenation since the packets are concatenated in one packet.
Node 1
Node 2
φ
Own data Node 1 Node 2
Own data
Figure 3.2: An example of aggregation in WBANs. The node receives data from nodes 1 and 2 with length 1. The header is removed, only the payload remains (length φ). This payload is combined with the node’s own data to form a new packet and a new header is added. The length of the packet is now 2φ+1. If no aggregation is used, there would be three packets with length 1.
CHAPTER3 53
3.1.3 Reliability
The influence of the network architecture on the reliability in WBANs was briefly studied in [10]. The star and multi-hop topologies are evaluated against the packet delivery ratio in an experimental setting. The packet delivery ratio could by in-creased by either operating at a higher power level or by changing to a multi-hop architecture. As the energy consumption is restricted, using a multi-hop topology is the preferred way to go.
Reliability has also been studied in wireless sensor networks. The authors of [11] claim that using the shortest path metric is not enough and propose a routing metric based on the link quality in order to increase the reliability of the network.
This metric is also used in [12] but has the tendency to generate long paths and it does not evaluate the energy consumption. A similar approach is used in [13] but tries to limit the number of retransmissions using a multi hop architecture. This leads to smaller and more energy efficient networks.
In this work, the reliability is studied for different metrics such as packet de-livery ratio using the link probability and collision of probability. The influence of the energy efficiency on the reliability is investigated and the ILP-formulation is once again extended so that the most reliable and energy efficient path is chosen.