Table 3.2 lists the parameters we use in the ns-2 simulations. We use the IEEE 802.11b wireless module with link rate of 11M bps. RTS/CTS hand-shaking is disabled. All nodes are uniformly deployed in an area of 220 × 220 sq. meters. As shown in Fig. 3.3, both single-hop and multi-hop traffic are generated for grid networks of 9, 25, and 49 nodes. To avoid the corner effect which may bias the results, we actually generate more nodes and traffic flows so that the corner nodes can have the same surroundings as the central nodes.
Simulation statistics are obtained from the central 9, 25, and 49 nodes of the network. In Fig. 3.4(a), Default indicates the method with no power con-trol (using default transmit power), whereas N-base means the method that applies N-base protocol. We observe that for single-hop traffic (Fig. 3.3(a)), N-base performs much better especially in dense networks. This is because more spatial diversity is achieved by N-base. Recall that n is the number of connected neighbors when using the default transmit power. According to the derivation in Eq. 3.9, the ratio of log nn determines the scale that system throughput can be improved by power control. In our experiments, n = 8, 24,
and 44 for the 9-, 25-, and 49-node grid topologies. Apparently the denomi-nator log n of Eq. 3.9 increases much slower than the numerator n as n grows.
Thus the ratio is expected to increase drastically as network becomes dense, which explains Fig. 3.4(a). Note that in our grid examples, due to the equal distance between four closest neighbors, in our simulations, the number of connected neighbors after N-base power control is always four. The reason is the logarithms of 9, 25, and 44 are all less than four, and in grid topology, a node will connect to zero neighbor if power is reduced to connect to less than four neighbors (i.e., log n = 4 for all three node densities). In order to vali-date our analytic model, we obtain the ratio of simulative system throughput with Default to that with N-base, and compare to the analytic ratio of log nn . The results in Fig. 3.4(b) shows that the analytic predictions are quite close to the simulative outcomes.
Table 3.2: Parameters used in NS-2 simulation Two-Ray Ground Propagation Model Antenna Gain = 1 CWmin = 32
CWmax= 1024 Default Ptr = 20dBm Noise = −101dBm CSThreshold = −99dBm RXThreshold = −95dBm
CBR sending rate = 1Mbps Packet size = 1024 bytes
So far, power control seems to yield better system throughput by bring-ing more spatial diversity (enablbring-ing multiple concurrent communications).
However, as shown in Fig. 3.5, the N-base method performs poorly for the multi-hop traffic in terms of system throughput. This erratic phenomenon
suggests that the spatial diversity advantage of power control no longer domi-nates the performance for multi-hop traffic. In contrast, complicate inter-hop interference and lengthened packet route affect the multi-hop performance in a bad way. Motivated by this observation, we seek to balance the pros and cons of power control for multi-hop traffic with the assistance of using multiple wireless radio channels.
(a) shows the interference power before and after power control.
(b) illustrates the ratio of default interference (transmit) power di-vided by reduced interference (transmit) power.
Figure 3.1: Impact of power control on interference power and transmit power.
Figure 3.2: Impact of power control on network capacity.
(a) Single-hop traffic
(b) Multi-hop traffic
Figure 3.3: The single-channel single-radio grid network with 9, 25, and 49 nodes respectively.
(a) Throughput
(b) Ratio of CCpmm
Figure 3.4: Single-hop traffic performance in a single-channel single-radio grid network.
Figure 3.5: System throughput for multi-hop traffic in a channel single-radio grid network.
Chapter 4
Multi-channel Multi-radio Grid Network
In this section, we consider a grid network with I radio interfaces at each node, running over C non-interfering channels. Here I ≤ C. In case I < C, a common subset (with size I) of C channels will be selected so that every node uses the same channel set to configure channels for its I radios. We are interested in improving the system performance with multi-hop communica-tions. To this end, we first propose our GradPC framework in Section 4.1, and then report the performance results via simulations in Section 4.2.
4.1 Gradational Power Control Protocol (GradPC)
The design rationale behind the GradPC protocol is to impose power gra-dations on radios equipped at each node, so as to provide flexibility of
bal-Figure 4.1: The GradPC procedure to impose power gradations on radios at each node.
ancing the contradicting factors, such as route length and spatial diversity, for multi-hop traffic performance. In the proposed GradPC framework, a base channel is designated to always use the default transmit power Ptr (no
power control on this radio). In this way, the route can be kept short, and network connectivity can be preserved despite performing power reductions on the other non-base radios. Define the neighbor table (set) established over base channel as Nbase, and n denotes the cardinality of set Nbase (size of neighbor nodes over base channel). Parameter n can be easily obtained by implementing heart-beat message (e.g., HELLO) exchanging mechanisms at each node. Consequently, nodes can estimate their respective n value by periodically exchanging HELLO messages over the base channel. In addition, the base channel is responsible for finding packet routes due to its high net-work connectivity. In the current GradPC framenet-work, we adopt a variant of DSR routing mechanism, which always gathers three possible routes and then randomly chooses one. In contrast to favoring the shortest route in default DSR, the selected route in our GradPC protocol may not be the shortest.
Generally speaking, the shortest route comes with longer traveling distance between hops. In order to support long transmitting distance, high transmit power should be used. As a result, we observe that in many cases, default transmit power is necessary to support the route discovered by default DSR over the base channel. On the other extreme, we may choose the longest route, which produces short traveling distance between hops. In this case, the required power level can be reduced, but the end-to-end throughput may suffer due to many unnecessary relays. The above observations motivate us to adapt the DSR protocol. Our objective is to determine a moderate route path which has mixed short and long hops. Such route provides us flexibility
of scheduling different channels and power levels to be used between hops.
For non-base radios, our GradPC adopts the N-base protocol as the power control mechanism. Specifically, once n is obtained from the base channel, the GradPC procedure reduces power levels gradationally so that the connec-tivity degrees for non-base channels become less and less. Fig. 4.1 illustrates the GradPC procedure. After GradPC procedure is done, the transmit power level Pti that should be used by radio i is obtained. Then each non-base radio should perform the heart-beat message exchanging function to establish the neighbor table (set) Nifor radio interface i. Note that when tuning the power level for a non-base radio, we follow the ns-2 settings which divides power into ten levels ranging from 1mW to 100mW. That is, power is reduced by 10mW at a time until the number of connected neighbors satisfies the de-sirable number. Once the power levels have been determined for all radios, and route is ready, an interface scheduling procedure is performed to sched-ule the next packet to be sent over an appropriate channel (radio). Given a packet route, we consider both channel diversity between hops and spatial reuse factor resulted from power control. Generally, the radio interface with the lowest transmit power is preferred, suppose the next hop is reachable using this transmit power. In addition, to provide channel diversity between hops, we propose to circulate the channel assignment by avoiding the chan-nel used by the previous hop. Define Chpre hop as the channel ID used by the previous hop. Each node sets the initial channel ID to be considered as fI(Chpre hop− 1), where fI is a circulation function, so that the function
value always takes on some integer between [1, I]. This mechanism does pro-vide certain channel diversity between hops, but do not guarantee absolute diversity. Fig. 4.2 illustrates our interface scheduling policy in the GradPC framework. More detailed pseudo-codes are presented below to show the internal operations of the GradPC protocol.
Figure 4.2: Interface scheduling policy after route is ready.
Algorithm 1 Establish neighbor table using base channel: estimation of parameter n
1: Pt ← Ptr // set default transmit power
2: n ← 0
3: Send message(HELLO) periodically over base channel using transmit power Pt
4: while (HELLO received) and (!Timeout) do
5: Add node’s ID to neighbor table Nbase
6: n ← n + 1
7: end while
Algorithm 2 GradPC procedure: power adaptation policy for respective radio interface at each node
1: I ← Number of interfaces
2: i ← 1 // interface index
3: a1 ← n // n obtained from Algorithm 1
4: while i ≤ I do
5: Pti ← P (ai) // power adjustment function for radio i to connect to ai neighbors
6: Establish neighbor table Ni
7: if ai ≥ e then
In this section, we extend the ns-2 code to support multi-channel multi-radio environment. We use the 3 non-overlapping channels (numbering as channel 1, 2, and 3) in IEEE 802.11b, and install 3 radio interfaces at each node.
Channel 1 is designated as the base channel. The same ns-2 parameters
Algorithm 3 Interface selection procedure: data will be sent over the se-lected radio
1: if First hop then
2: i ← I // initial interface index
3: else
4: i ← fI(Chpre hop− 1)
5: end if
6: while i ≥ 1 do
7: if Next hop found in Ni then
8: Data sent over radio i
9: else
10: i = i − 1
11: end if
12: end while
// next hop unreachable
13: Re-discover route on base channel
(Table 3.2) and network topologies (Fig. 3.3) are used in our simulations.
We investigate the system throughput of multi-hop flows (Fig. 3.3(b)) for three approaches: GradPC, N-base, and Default. All three approaches use 3 non-overlapping channels and 3 radio interfaces at each node. Default indicates the method of using default transmit power for all three radios, whereas N-base denotes the approach of applying the same power level to connect to log n neighbors for all three radios. Since there is no interface scheduling mechanism specified for Default and N-base, in order not to take advantage of them in this regard, we implement the same interface scheduling algorithm (shown in Fig. 4.2) as GradPC in Default and N-base. For routing strategy, Default and N-base use the shortest routes found by DSR using their respective power levels, while GradPC use routes randomly chosen from the
first three routes discovered by DSR (explained previously in Section 4.1).
Fig. 4.3(a) plots the system throughput of multi-hop traffic flows (gen-erated as in Fig. 3.3(b)). With the assistance of channel diversity, the formance of Default and N-base is comparable, in contrast to the sharp per-formance degradation produced by N-base as previously shown in Fig. 3.5 when C = 1 (single-channel environment). From Fig. 4.3(a), we observe that our GradPC performs the best especially for dense networks. To get a better understanding of the impact on multi-hop traffic performance, we give an-other set of statistics in Fig. 4.3(b), which shows the system performance of a dense grid network (49 nodes) as the number of multi-hop flows increases.
As we can see from this figure, when C = 1 (single-channel system), no power control is suggested in terms of better multi-hop traffic performance. When C = 3 (multi-hop environment), interestingly, N-base is not always worse than Default. For environments with very light and very heavy loads (2 and 7 flows), N-base even performs better than Default. We extrapolate from the results that both route length and medium utilization (spatial diversity) play an important role for multi-hop traffic performance. Our GradPC out-performs other mechanisms in all cases especially when traffic load is heavy (7 flows).
Table 4.1 summarizes the hop count information for the three methods.
Our GradPC uses the routes with moderate lengths (neither the shortest nor the longest) in order to preserve both the advantage of power control (increased spatial reuse factor) and channel diversity (decreased inter-hop
interference), hence explains the good performance in Fig. 4.3.
Table 4.1: Hop count statistics in a 49-node grid network GradPC N-base Default
Total hops 28 42 14
Avg. hops 4 6 2
(a) C = 3
(b) 49 nodes
Figure 4.3: Multi-hop traffic performance in a multi-channel multi-radio grid network.
Chapter 5
Applying GradPC in
Multi-channel Multi-radio Random Topology
We set up a multi-channel multi-radio network with 50 nodes randomly de-ployed and randomly generate 7 multi-hop flows, as shown in Fig. 5.1(a).
Three 802.11b non-overlapping channels are used. The three network topolo-gies produced by our GradPC are illustrated in Fig. 5.1(b)(c)(d) respectively.
One more method, BICONN, is implemented for providing another power control alternative besides N-base. The BICONN protocol is a power control mechanism proposed by [19]. With multiple channels, BICONN applies the same power reduction for all radios (as the N-base does). We create CBR traffic and increase the sending rate to 11M bps. Fig. 5.2 shows the multi-hop
(a) 7 data flows (b) topology l
(c) topology 2 (d) topology 3
Figure 5.1: Illustration of node and flow distributions, along with the con-nected network topologies using GradPC over three channels.
system throughput for different methods as simulation time advances. From this figure, we observe that our GradPC outperforms other methods, and has the highest saturated throughput. Table 5.1 provides the hop count infor-mation for all methods. In this case, our GradPC happens to have the same hop count as Default. Nonetheless, since GradPC imposes power gradations on radios, while Default applies the same default transmit power (without
power reduction) for all radios, GradPC still yields much better performance than Default, due to higher spatial reuse factor. Moreover, Default is even worse than both N-base and BICONN.
Figure 5.2: Performance comparison of multi-hop traffic in a 50-node random network topology with 7 flows.
Combining all the previous results from both grid and random network topologies, we demonstrate that multi-hop system performance cannot be determined by power parameter or route length alone. Instead, factors such as power, channel, and routing strategy all co-dominate the system perfor-mance of multi-hop flows. By seeking tradeoff between those factors, our proposed GradPC framework helps open up more system capacity for multi-hop communications.
Table 5.1: Hop count statistics in a 50-node random network topology GradPC BICONN N-base Default
Total hops 21 30 26 21
Avg. hops 3 4.285 3.714 3
Chapter 6
Conclusion and Future Work
In this paper, we did a pilot study on the interaction of two physical pa-rameters: power and channel, with the goal of further expanding the system throughput of multi-hop traffic in a wireless ad hoc network. We proposed GradPC and its accompanying route and channel selection protocols. In the current proposal, we adopted the N-base protocol as our power control mech-anism to provide the power gradations over radios. However, one may cus-tomize other existing power control strategies in place of the N-base protocol.
In addition, though the cost of wireless cards has become quite affordable, in some cases it is difficult to install multiple radios at a computing device, due to size consideration or hardware support availability. Thus, how to utilize multiple channels based on the GradPC concept by practically using a single radio may be worth future investigation. This becomes challenging because, in this case, we should carefully deal with both the switching issues and
multi-channel hidden-terminal problem, inevitably at the cost of significant control signaling overhead.
The simulation results showed that the proposed GradPC protocol suite yielded the most system throughput than other strategies (including strategy with no power control, and strategies keeping the same channel connectivity degree). By imposing power gradations on radios, and considering route and channel scheduling, our proposed techniques have effectively balanced the multi-hop performance requirements for shorter route and better spatial diversity.
Bibliography
[1] A. Akella, G. Judd, S. Seshan, and P. Steenkiste. Self Management in Chaotic Wireless Deployments. In Proc. ACM MobiCom, 2005.
[2] D. M. Blough, M. Leoncini, G. Resta, and P. Santi. Topology Con-trol with Better Radio Models: Implications for Energy and Multi-hop Interference. In Proc. ACM MSWiM, 2005.
[3] D. M. Blough, M. Leoncini, G. Resta, and P. Santi. The k-Neighbors Approach to Interference Bounded and Symmetric Topology Control in Ad Hoc Networks. IEEE Transactions on Mobile Computing, 5(9):1267–
1282, 2006.
[4] B. Bollob´as. Random Graphs. Academic Press, Orlando, FL, 1985.
[5] M. Burkhart, P. Rickenbach, R. Wattenhofer, and A. Zollinger. Does Topology Control Reduce Interference? In Proc. ACM MobiHoc, 2004.
[6] F. Cali, M. Conti, and E. Gregori. Dynamic Tuning of the IEEE 802.11 Protocol to Achieve a Theoretical Throughput Limit. IEEE/ACM Transactions on Networking, 8(6):785–799, December 2000.
[7] X. Guo, S. Roy, and W. S. Conner. Spatial Reuse in Wireless Ad-hoc Networks. In Proc. IEEE VTC, 2003.
[8] P. Gupta and P. R. Kumar. Critical Power for Asymptotic Connectivity in Wireless Networks. Stochastic Analysis, Control, Optimization and Applications, 1998.
[9] P. Gupta and P. R. Kumar. The Capacity of Wireless Networks. IEEE Transactions on Information Theory, 46(2), March, 2000.
[10] E. S. Jung and N. H. Vaidya. A Power Control MAC Protocol for Ad Hoc Networks. In Proc. ACM MobiCom, 2002.
[11] T.-S. Kim, H. Lim, and J. C. Hou. Improving Spatial Reuse through Tuning Transmit Power, Carrier Sense Threshold, and Data Rate in Multihop Wireless Networks. In Proc. ACM MobiCom, September 2006.
[12] P. Kyasanur and N. H. Vaidya. Routing and Interface Assignment in Multi-channel Multi-interface Wireless Networks. In Proc. IEEE WCNC, March 2005.
[13] J. Li, C. Blake, D. S. J. De Couto, H. I. Lee, and R. Morris. Capacity of Ad Hoc Wireless Networks. In Proc. ACM MobiCom, July 2001.
[14] N. Li and J. C. Hou. Topology Control in Heterogeneous Wireless Net-works: Problems and Solutions. In Proc. IEEE INFOCOM, 2004.
[15] T.-Y. Lin and J. C. Hou. Interplay of Spatial Reuse and SINR-determined Data Rates in CSMA/CA-based, Multi-hop, Multi-rate Wireless Networks. In Proc. IEEE INFOCOM, 2007.
[16] J. P. Monks, V. Bharghavan, and W. W. Hwu. A Power Controlled Multiple Access Protocol for Wireless Packet Networks. In Proc. IEEE INFOCOM, 2001.
[17] A. Muqattash and M. Krunz. Power Controlled Dual Channel (PCDC) Medium Access Protocol for Wireless Ad Hoc Networks. In Proc. IEEE INFOCOM, 2003.
[18] A. Muqattash and M. Krunz. A Single-channel Solution for Transmission Power Control in Wireless Ad Hoc Networks. In Proc. ACM MobiHoc, May 2004.
[19] R. Ramanathan and R. Rosales-Hain. Topology Control of Multihop Wireless Networks using Transmit Power Adjustment. In Proc. IEEE INFOCOM, 2000.
[20] P. Santi. Topology Control in Wireless Ad Hoc and Sensor Networks.
ACM Computing Surveys (CSUR), 37(2):164–194, June, 2005.
[21] Y.-C. Tseng, S.-L. Wu, C.-Y. Lin, and J.-P. Sheu. A Multi-channel MAC Protocol with Power Control for Multi-hop Mobile Ad Hoc Net-works. In Proc. IEEE Int’l Conference on Distributed Computing Sys-tems (ICDCS), pages 419–424, 2001.
[22] F. Xue and P. R. Kumar. The Number of Neighbors Needed for Con-nectivity of Wireless Networks. Wireless Networks, 10:169–181, 2004.
[23] X. Yang and N. Vaidya. On Physical Carrier Sensing in Wireless Ad Hoc Networks. In Proc. IEEE INFOCOM, April 2006.