Simulation Results of Spiral Grid Routing Protocol

In this section, we evaluate the performance of our proposed scheme. Our experimental results are based on a full implementation of the previously described GGR and LBDD protocols. We focus on the comparison of load balance feature among our proposed SGR, GGR, and LBDD protocols.

We focus on the load balance feature of energy consumption. The performance metrics include the remaining energy after executing several queries and the nodes death rate after collecting several rounds of data.

For evaluating the remaining energy, we use the standard deviation among nodes to evaluate the performance of load balance. That is, the smaller the standard deviation of remaining energy is, the better performance in load balance has achieved.

The standard deviation Std is calculated by using the following equation:

𝑆𝑡𝑑 = (𝑥_𝑖 − 𝑥 )² 𝑁 − 1

Where x_i is the remaining energy for node i, 𝑥 is the average energy remaining of all sensors and N is the number of sensors in the field.

In this simulation, we assume that the sink node has 50 queries to be sent to the region of interest and each query will collect data from the region of interest for 20 times. The grid shifting will be launch after collecting 10 times of data. Figure 4.1 shows the simulation result. As we can see in Figure 4.1, our proposed SGR protocol outperforms than GGR and LBDD protocols in the balance of remaining energy. This means that our SGR protocol can distribute the work of energy consumption more evenly to other nodes in the grid cell. Especially, as the number of queries been

33

executed grows, the load balance feature in energy consumption shows that our SGR protocol performs much better than the other protocols.

Figure 4.1. Standard deviation of remaining energy.

For evaluating the node death rate, we are interesting in knowing the percentage of nodes which are out of energy after executing queries. If the percentage is low, then the lifetime of the network can be prolonged. In this simulation, we evaluate the node death rate by executing a query sent from sink. Once a sensor within the region of interest received the query message, it will send back the observation data to the sink periodically. For convenience, we call it a round after the sensor node has sent back 20 times of data. Note that we only compare our proposed SGR protocol with LBDD protocol. The reason is that, as the result from Figure 4.1, SGR and LBDD protocols are more load balance than GGR. Therefore, we focus on the comparison of these two protocols to understand which one can make the network’s lifetime longer.

Figure 8 shows the simulation result. In Figure 4.2, the node death rate of our SGR protocol is much lower than LBDD when the number of rounds is less than 100. Our SGR protocol will obtain the same node death rate as LBDD after 160 rounds. This means that our SGR protocol can effectively prolong the lifetime of the network.

0 200 400 600 800 1000 1200 1400 1600

10 20 30 40 50

Standard deviation of remaining energy (Jules)

Number of events been excuted

SGR LBDD GGR

34

Figure 4.2. Node death rate vs. number of rounds.

For evaluating the effectiveness of our grid shifting, we are interested in knowing the node death rate after performing several grid shifting processes. In this simulation we compare our SGR protocol only with the GGR protocol. This is because that our SGR and LBDD protocols are taking the similar mechanism to change the grid structure in which the cell size is fixed. These are different from the mechanism taken by GGR in which the cell size is changed in order to change the grid structure. Figure 9 shows the simulation result. In this simulation, we compare the node death rate of our SGR with three different cell sizes of GGR protocol, which are fixed size, large size and small size. The fixed size of GGR in the simulation means that the cell size is fixed during the simulation. The large size means that the cell sizes are incremented from initial cell size, say , to 1.1, 1.2, …, and 1.9, and then return to  and increment the cell size as previous again. The small size means that the cell sizes are decremented from initial cell size to 0.9, 0.8, …, and 0.1, and then return to  and decrements the cell size as previous again. In Figure 4.3, the node death rate of our SGR protocol is lower than all three cell sizes of GGR protocol.

Especially, the node death rate of the small size approach of GGR is much larger than

0 1 2 3 4 5 6 7 8 9 10

20 40 60 80 100 120 140 160

Node Death Rate(%)

Number of Rounds

SGR LBDD

35

the others after executing more than 60 grid shifting or changing processes. This is because that more nodes are involved in the data forwarding process when the grid size is small. Therefore, more nodes will be out of energy in this situation. Finally, our SGR protocol will obtain the smaller node death rate than GGRs, indicating that our grid shifting mechanism performs more effective than the grid changing mechanism in GGR protocol.

Figure 4.3. Node death rate vs. number of grid shifting.

0 10 20 30 40 50 60 70 80

10 20 30 40 50 60 70 80

Node Death Rate(%)

Number of Grid Shifting

SGR GGR(fixed) GGR(big) GGR(small)

36