E-textile Based Imaging Array: Final Phase

Introduction

Comparison of Routing Algorithms

Performance
Table 1 shows the time taken for the individual nodes to complete the ID map forming step for both flooding and ring algorithms. Not all the nodes in the flooding case were able to complete the ID map before the batteries for the nodes are depleted. Clearly, the flooding algorithm is not suitable.

Figure 1 shows the cumulative time taken for the number of nodes that complete the ID map.

Flooding		Ring
Node ID	Time taken (sec)	Node ID	Time taken (sec)
14	0.411443	8	0.172628
13	0.421562	13	0.19577
19	0.49638	9	0.195778
18	0.503121	12	0.204218
12	0.551108	14	0.205012
7	0.724927	17	0.228353
11	0.727103	19	0.230589
8	0.758391	7	0.240746
6	0.775167	18	0.263761
2	0.826725	10	0.289727
20	0.827247	24	0.290753
9	0.839707	16	0.290804
1	0.852532	15	0.294659
3	0.971233	23	0.295735
17	0.98721	11	0.297355
22	1.094035	6	0.319901
23	1.415957	2	0.347763
24	2.010461	3	0.368582
		20	0.37821
		25	0.420266
		22	0.428763
		5	0.42998
		4	0.448707
		1	0.470458
		21	0.55273

Figure 1 Time taken to form ID map (flooding vs ring)

Power
Figures 2 and 3 shows the battery current utilization for one of the nodes (Node 1) for the two routing algorithms respectively. Although the difference is not very significant, it can be seen that there are more high instantaneous values for the flooding case. Thus in terms of power utilization, the ring algorithm would also be a better option.

Figure 2 Battery current for flooding algorithm

Figure 3 Battery current for ring algorithm

Ring vs. Fixed Routing Algorithm
The ring algorithm has also been compared against the fixed routing algorithm in terms of performance and power utilization. The ring algorithm has been used for the ID map formation step (Step 2) in both cases. The fixed algorithm is only used for the image forming step (Step 3). Thus the two cases for comparison are “ring-ring” and “ring-fixed”, as labeled in the diagrams.

Performance
The “quality” of the image formed is used as a comparison for performance. Samples of the image are collected at periodic intervals. A value for a node in the image sample represents the absolute time that the sensor data was produced. There are two criterion used to compare performance. The first is “image coherence” which is the standard deviation of the times for the 25 nodes in the image. The second measure is the “image timeliness” which is the difference between the value for Node 13 (the collector for the sensor data from all nodes) and the mean of the values of all the nodes in the image. Node 13 will always be the most updated since it is forming the image.

Note that a “coherent” image may not necessarily be the most updated.

Figures 4 and 5 shows the “image coherence” and “image timeliness” respectively. As can be seen, the fixed routing algorithm has performed better than the ring algorithm, both in terms of coherence and timeliness.

Figure 4 Image coherence

Figure 5 Image timeliness

Power
Figure 6 shows the battery drainage for the both cases. The data obtained represents the sum of the battery remain energies for all 25 nodes in the array. Although the difference is not significant, it can been seen that the ring algorithm is draining battery energy faster than the fixed algorithm.

Figure 6 Battery energy utilization

Conclusion

From this project, we can see that there are many implementation considerations when trying to implement an e-textile based imaging array. One of the main considerations is the routing algorithm used to form the ID map and the sensor image. The results showed that the quality of the image formed may not be very high, as a result of network latencies. Without efficient routing algorithms, both in terms of performance and power consumption, such e-textile based imaging array may be limited to lower end applications.

To achieve power efficiency and fault tolerance required by the e-textile based imaging array, more work need to be done to optimize the communications protocol and routing algorithms. This is particularly important when the imaging array is increased in size.

References