Chapter 6 Simulation Setup
6.2 Environment Setting in NS3
The simulation of model accuracy is performed in the end-to-end transmission in NS3. The Environment Setting in NS3 is shown in table 6-1. In our simulation, we employ the User Datagram Protocol (UDP) and Transmission Control Protocol (TCP) in
Path Loss (dB)
Time (s)
• Proposed Walking Model
• Measured Data
• Gamma Walking model
Time (s)
Path Loss (dB)
• Measured Data
• Proposed Sleeping Model
32
transport layer. UDP assumes that error checking and correction is either not necessary or performed in the application. In such case, there is no queuing process in upper layer, and the simulation results can completely reflect the queuing state in MAC layer. On the other hand, TCP provides reliable, ordered delivery of a stream of bytes from a program on one computer to another program on another computer.
The simulation with TCP as the transport layer will further demonstrate the correctness of our channel model in end-to-end reliable transmission. The accuracy is evaluated with end-to-end delay difference and throughput. In this thesis, the delay difference is defined as the difference of delay between the consecutive packets.
Table 6-1 Environment setting in NS3 simulation
Topology point-to-point
Packet size 250bytes
Traffic loading Uplink 50kbps
PHY speed 6Mbps
Tx Power -10dbm
Access control CSMA/CA
Transport layer TCP, UDP
Evaluation end-to-end delay difference, throughput
33
Chapter 7 Simulation Results
In this chapter, we show the simulation result for with measured data and our channel model to see the modeling accuracy of the proposed dynamic channel model – two-state walking model and three-level sleeping model. The UDP and TCP are employed in transport layer. We will focus on two factors such as end-to-end delay difference and throughput in end-to-end transmission, and we analyse the statistical distribution of them to compare with the measured data. We use those results to demonstrate the proposed dynamic channel model can truly reflect the queuing state in both MAC layer and end-to-end reliable transmission.
7.1 Simulation of two-state walking channel model 7.1.1 UDP Analysis
7.1.1.1 Throughput
Figure 7-1 Throughput in UDP analysis - Walking model
Throughput (kbps)
Tx Time(s)
Measured Data
Two-state model
Gamma model
34
Figure 7-2 Distribution of throughput in UDP analysis - Walking model
Table 7-1 Statistic of throughput in UDP analysis - Walking model Mean of throughput(kbps) σ of throughput(kbps)
RawData 38.53 18.91
TwoState 44.11 11.66
Gamma Model 48.67 7.43
Throughput (kbps)
density
35
7.1.1.2 Delay Difference
Figure 7-3 Delay difference in UDP analysis - Walking model
Figure 7-4 Distribution of delay difference in TCP analysis - Walking model Tx time(s)
Measured Data
Two-state model
Gamma model
Delay Difference(ms)
Delay difference(ms)
density
36
Table 7-2 Statistic and packet error rate of delay difference in UDP analysis - Walking model
7.1.1.3 Discussion of UDP analysis
Because of the error checking and correction is not performed in UDP, MAC layer is in charge of the whole retransmission processes. It retransmits the error packet until the next packet enters from upper layer. Therefore, the throughput of UDP analysis would not exceed the maximum traffic loading, 50kbps.
The throughput result shows our model can truly reflect the channel condition in MAC, high throughput in high-state, otherwise, low throughput in low-state.
Otherwise, the NICTA’s channel model can’t reply this condition, because it doesn’t consider the time-domain correlation in model.
However, as shown in Table 7-2, UDP has incomplete delay information due to packet dropping. The deviations of delay difference for those three path loss data are very small. Thus, both two-state and gamma model cannot correctly reflect statistic of delay difference.
37
7.1.2 TCP Analysis 7.1.2.1 Throughput
Figure 7-5 Throughput in TCP analysis - Walking model
Figure 7-6 Distribution of throughput in TCP analysis - Walking model Tx time (s)
Measured Data
Throughput (kbps)
Two-state model
Gamma model
Throughput (kbps)
density
38
Table 7-3 Statistic of throughput in TCP analysis - Walking model Mean of throughput (kbps) σ of throughput(kbps)
RawData 49.02 88.54
TwoState 48.96 78.12
Gamma Rnd 48.93 15.95
7.1.2.2 Delay Difference
Figure 7-7 Delay difference in TCP analysis - Walking model Tx time(s)
Measured Data
Two-state model
Gamma model
Delay Difference(ms)
39
Figure 7-8 Distribution of delay difference in TCP analysis - Walking model
Table 7-4 Statistic of delay difference in TCP analysis - Walking model Mean of delay difference (ms) σ of delay difference (ms)
RawData 0.04 83.64
TwoState -0.24 85.65
Gamma Rnd -0.002 21.73
7.1.2.3 Discussion of TCP Analysis
Due to the reliable transmission characteristic of TCP, it would persistently retransmit those packets in queue. Thus, lead to the throughput as shown in Figure 7-5, which the low RSSI of measured data bring about packets start to queue and cause low throughput, and high RSSI value leads to high success rate and high throughput in the end-to-end transmission. And the delay difference as shown in
Delay difference(ms)
density
40
Figure 7-7 also performs the same feature we just mentioned above. The low RSSI causes bad transmission quality with high delay difference, and high RSSI leads to no delay difference in end-to-end connections. By grasping this time-domain characteristic in building a channel model, our two-state can well reflect this queuing phenomenon. On the contrary, the Gamma model which doesn’t consider the time-domain correlation shows poor modeling accuracy with measured data.
7.2 Simulation of Three-level Sleeping channel model 7.2.1 UDP Analysis
7.2.1.1 Throughput
Figure 7-9 Throughput in UDP analysis - Sleeping model
Tx Time(s)
Throughput (kbps)
Measured Data
Three-level model
41
Figure 7-10 Distribution of throughput in UDP analysis - Sleeping model
Table 7-5 Statistic of throughput in UDP analysis - Sleeping model Mean of throughput(kbps) σ of throughput(kbps)
RawData 46.91 11.60
Three-Level 47.54 9.81
Throughput(kbps)
density
42
7.2.1.2 Delay Difference
Figure 7-11 Delay difference in UDP analysis - Sleeping model
Figure 7-12 Distribution of delay difference in UDP analysis - Sleeping model Tx time(s)
Delay Difference (ms)
Measured Data
Three-level model
Delay Difference (ms)
density
43
Table 7-6 Statistic and packet error rate of delay difference in UDP analysis - Sleeping model
7.2.1.3 Discussion of UDP analysis
As we mentioned in chapter 6, the RSSI of measured data can be classified to three-level, 1) 100% probability to get maximum throughput; 2) 50% probability to reach the highest throughput; 3) no chance to correctly transmit packet; after the retransmission process in MAC. In Figure 7-9, the throughput for measured data in UDP analysis also supports this assumption. The RSSI which is higher than the RSSIhigh −threshold can get the maximum throughput, 50kbps. And The RSSI which is higher than the RSSIlow −threshold and lower than the RSSIhigh −threshold only has 50% to reach the highest throughput. Besides, the packet which the RSSI is too low to successfully retransmit gets no throughput after retransmission in MAC. Our three-level sleeping model which considers this feature in model can get the perfect match to measured data in throughput. On the other hand, due to the incomplete delay information of dropped packets, the deviations of delay difference for those two path loss data are very small. Thus, our proposed model cannot wholly reflect statistic of delay difference.
44
7.2.2 TCP Analysis 7.2.2.1 Throughput
Figure 7-13 Throughput in UDP analysis - Sleeping model
Figure 7-14 Distribution of Throughput in TCP analysis - Sleeping model Tx time (s)
Throughput (kbps)
Measured Data
Three-level model
Throughput(kbps)
density
45
Table 7-7 Statistic of throughput in TCP analysis - Sleeping model Mean of throughput(kbps) σ of throughput(kbps)
RawData 49.94 71.75
Three Level 49.99 61.94
7.2.2.2 Delay Difference
Figure 7-15 Delay difference in TCP analysis - Sleeping model
Tx time(s)
Delay Difference (ms)
Measured Data
Three-level model
46
Figure 7-16 Distribution of delay difference in TCP analysis - Sleeping model
Table 7-8 Statistic of delay difference in TCP analysis - Sleeping model Mean of throughput σ of throughput
RawData 0.57 768.20
Three Level ~=0 (-8.27e-007) 625.22
7.2.2.3 Discussion of TCP analysis
As we mentioned before, TCP is a reliable transmission protocol, it would persistently retransmit those packets in queue. If we consider the time-domain correlation and three-level characteristic in model, we can get a better match to the measured data. The statistics of throughput and delay difference are shown in Table 7-7 and 7-8, respectively. The statistics demonstrate our three-level sleeping model has high modeling accuracy in TCP analysis.
Delay difference(ms)
density
47
Chapter 8
Conclusion and Future Work
8.1 Conclusion
In this thesis, we propose the two-state walking model and three-level sleeping channel model for dynamic body channel measurements and the modeling accuracy of those two models have been demonstrated in NS3 simulations. Using the concept of time-domain correlation, our proposed models can truly reflect the queue state in MAC. In walking channel analysis, we find the RSSI in walking can be properly modeled by two-state walking model, which is corresponding to the walking process.
The duration of each walking period and the ratio of upper and lower state duration can be modeled by lognormal distribution. And the received power amplitude of upper and lower state can be well characterized by lognormal distribution. On the other hand, the three-level sleeping model has been proposed. After we discover the
“stair feature” of measured sleeping channel, we combine this feature with the retransmission characteristic in MAC to classify the RSSI value to three levels. The three-level model is composed of many short states, and the duration of each state can be well characterized by exponential distribution. The correlation between consecutive levels is modeled by Markov chain. The simulation of throughput and delay difference evaluation in both UDP and TCP analysis demonstrate our proposed two-state walking model can be a better fit as we compare with the existent walking model proposed by NICTA. Besides, we have also proved our proposed three-level sleeping model can well catch the channel characteristics of this vital behavior in MAC as we consider both the time-domain and RSSI correlations in our design.
48
8.2 Future Work
The proposed channel models will be more accurate than others in MAC point of
view. But it’s still not enough to model all the activities of everyday life. In the future, we will try to model other scenarios, like driving, running, etc. Besides, reliability issue is one of the most important considerations in health-care applications of WBAN. And low power is another key design challenge of WBAN, which is essential for long-duration measurement in many WBAN medical applications. We should extend our channel model to accomplish both of the two important issues in the future work.
49 Wireless Body Area Network with Frequency Pre-Calibration," in Proc. 2008 IEEE VLSI-DAT, pp. 192-195, Apr. 2008.
[3] E. Reusens, W. Joseph, G. Vermeeren, D. Kurup and L. Martens, "Real human body measurements, model, and simulations of a 2.45 GHz wireless body area network communication channel," in Medical Devices and Biosensors, 2008.
ISSS-MDBS 2008. 5th International Summer School and Symposium on, 2008, pp. 149-152.
[4] J. Ryckaert, P. De Doncker, R. Meys, A. de Le Hoye and S. Donnay, "Channel model for wireless communication around human body," Electron. Lett., vol.
40, pp. 543-544, 2004.
[5] K. Takizawa, A. Aoyagi, J. Takada, N. Katayama, K. Yekeh, Y. Takehiko and K. R.
Kohno, "Channel models for wireless body area networks," Conf. Proc. IEEE Eng. Med. Biol. Soc., vol. 2008, pp. 1549-1552, 2008.
[6] J. M. Choi, H. J. Kang and Y. S. Choi, "A study on the wireless body area network applications and channel models," in Future Generation Communication and Networking, 2008. FGCN'08. Second International Conference on, 2008,
[7] D. Smith, L. Hanlen, D. Miniutti, J. Zhang, D. Rodda and B. Gilbert, "Statistical characterization of the dynamic narrowband body area channel," in Applied Sciences on Biomedical and Communication Technologies, 2008. ISABEL'08.
50
First International Symposium on, 2008, pp. 1-5.
[8] D. Miniutti, L. Hanlen, D. Smith, A. Zhang, D. Lewis, D. Rodda and B. Gilbert,
[11] S. H. Cheng and C. Y. Huang, "Power model for wireless body area network," in IEEE Biomedical Circuits and Systems Conference, 2008. BioCAS 2008, 2008, pp.
1-4.
[12] National Chiao Tung University, Hsinchu, Taiwan," u-PHI Wireless Body Area Network Core Technology," Ministry of Economic Affairs academic in industrial technology development plan, July 2006
[13] H. Akaike," Information theory as an extension of the maximum likelihood principle, " in Proc. 2nd. Int. Inf. Theory Syst., 1973, pp. 267–281.
[14] A. Fort, C. Desset, P. deDoncker, P. Wambacq, and L. van Biesen, “An ultra-wideband body area propagation channel model-from statistics to implementation,” IEEE Transactions on Microwave Theory and Techniques, vol.
54, no. 4, pp. 1820–1826, April 2006.