This section reviews the main characteristics of simulation scenarios and the methods of collecting and analysing the readings of these scenarios. In general, this thesis considers a number of sensor nodes that have identical equipment (radio transceivers, battery capacity, etc.) and are distributed uniformly over a given area to form a multihop network. The border effect is avoided by allowing nodes that are near the edge of the simulated area to communicate with the other nodes on the opposite sides. All the transmissions and receptions from nodes are performed with the same
M. Baz, PhD Thesis, University of York 2014
and fixed data rate and using the same power level that is obtained from a real data sheet [137].
This research employs an ideal propagation model with free space parameters. The main reason for this selection is that all the proposed protocols operate at the MAC and routing layers and hence incorporating a non-ideal model can hinder the discrimination between the effects of the proposed model and the effects of the randomness of the propagation model. Another reason for using an ideal propagation model is that WSNs are deployed over wide range of environments including indoor, outdoor and underwater. Since each one of these environments has its own characteristic that differ substantial than others, considering a particular propagation model can limit the applicability of the proposed protocols.
The physical parameters of the network (e.g., data rate, channel assessment mechanism and slot time) are adjusted according to the IEEE 802.15.4 physical specifications [16]. IEEE 802.15.4 is the official standard for low power wireless personal area networks which covers a wide communication systems and this standard has become an essential component of higher-level protocol stacks such as ZigBee [111] and 6LowPAN [17]. Hence adopting the physical layer parameters of the IEEE 802.15.4 standard in this thesis can widen the applicability of the proposed protocols and make them ready to be integrated with future standards.
Each simulation scenario is typified by four parameters 〈𝑘, 𝑁, 𝑋, 𝑌, 𝑍〉 where the first two parameters (i.e., 𝑘, 𝑁) are used to characterise the topology of network, the second two parameters (i.e., 𝑋, 𝑌) represent the traffic patterns and finally the 𝑍 parameter defines the routing policy.
The parameter 𝑘 defines the connectivity degree of the network, i.e., the average number of mutually independent paths between non-neighbouring nodes. The parameter 𝑁 defines the average number of nodes per a channel which is the product of the average number of nodes per a square metre by the average area of the transmission range. The main reason for specifying this parameter is to provide a meaningful quantity that describes both the topology and transceiver parameters. The values of 𝑁 and 𝑘 are set by adjusting the transmission range and the number of nodes per unit area in order to generate a variety of network topologies [138].
The parameter 𝑋 defines the probability distribution that is used by nodes to generate their internal inter-arrival times between packets and the parameter 𝑌 specifies the rate of this traffic. This thesis uses three inter-arrival distributions, namely: Constant Bit Rate (CBR), Exponential (EXP) and Weibull (WBL). CBR represents the case when the intervals between consecutive packets are of a fixed length equals to the reciprocal of the traffic rate specified by the parameter 𝑌. In the EXP case, the intervals between consecutive packets are random variables following exponential distribution whose rate is defined by the value of parameter 𝑌. Finally, WBL represents the case when the intervals between consecutive packets are generated
M. Baz, PhD Thesis, University of York 2014
according to the Weibull distribution. The Weibull distribution is characterised by two parameters: shape and rate. The shape parameter is adjusted to low value (e.g. less than one) in order to obtain a heavy-tailed distribution, while the rate parameter is set to the value of parameter 𝑌. The main reason for using these three inter-arrival distributions is to assess the performance of the proposed protocol under different traffic patterns. CBR is suitable for the periodic reporting applications while EXP accounts for the event driven applications and finally WBL represents the heavy tailed traffic patterns found in some networks such as the Internet of Things [9]. The length of packets that is generated by all of these distributions are set at the Maximum Transmission Unit (MTU) of the IEEE 802.15.4 standard which is 133 bytes unless other values are specified in a particular assessments.
Finally, the parameter 𝑍 defines the routing policy. This study employs two policies. The first policy distributes traffic evenly amongst all other nodes within the network. This policy is denoted by 𝑈 and used to ensure that each node acts as a source, router and final destination simultaneously. The second traffic distribution policy is denoted by A and refers the case when the traffic that is generated by all nodes is destined to a single receiver selected randomly. This traffic distribution mode has been selected as it is suitable for cluster based WSNs. In chapters 6-7, the DSDV [105] routing protocol is used while chapter 8 proposes a novel routing protocol.
For each simulated scenario (tuple) fifty random topologies are generated and for each topology, the simulation sessions are repeated one hundred times each elapsing for 109 seconds. The readings of each scenario are averaged over all nodes within the networks and then the 95% confidence level with respect to the mean is considered in the results.
Other parameters that are closely related to the proposed protocols such as contention parameters of the CSMA-CA protocol or the routing metrics are stated in the corresponding chapters.
4-5 Conclusion
This chapter has reviewed the simulation techniques and result validation methods that will be used in this thesis. A review of some of the widely used simulation platforms demonstrates that the NS3 simulator offers appealing features which make it an excellent candidate to assess the performance of the proposed protocols in this thesis. This chapter has also provided an overview for the main characteristics of simulation scenarios and highlighted the methods of collecting and analysing the simulation outcomes. It can be seen from this overview that these configurations account for different cases which facilitates predicting the performance of the proposed protocols under wide variety of configurations.
M. Baz, PhD Thesis, University of York 2014