**1. Introduction**

102 Wireless Mesh Networks – Efficient Link Scheduling, Channel Assignment and Network Planning Strategies

Springer, ISBN 978-0-387-68839-8, New York

Layer Approach, *in IEEE WCNC*, 2005

ITU-T (2003), ITU-T Recommendation G.114, 2003.

2005, pp. 381- 390, 3-7 Oct. 2005

*Proceedings*, pp.1-12, April 2006

6, June 2011

19 Oct. 2010

*MC2R*, vol. 8, no. 2, pp. 50-65, April 2004

Networks, Proc. *ACM MobiCom*, pp. 114-128, 2004

*Information Theory*, volume 46, pp. 388–404, March 2000

Conti, M.; Das, S. K.; Lenzini, L. & Skalli H. (2007). Channel Assignment Strategies for Wireless Mesh Networks, In *Wireless Mesh Networks – Archi-tectures and Protocols*,

Draves, R.; Padhye, J. & Zill B. (2004). Routing in Multi-radio, Multi-hop Wireless Mesh

Gong, M. X. & Midkiff, S. F. (2005). Distributed Channel Assignment Protocols: A Cross-

Gupta, P. & Kumar, P. (2000). The Capacity of Wireless Networks, In *IEEE Transactions on* 

Husnain Mansoor Ali; Anthony Busson & Véronique Vèque (2009). Channel assignment algorithms: a comparison of graph based heuristics, In: *Proceedings of the 4th ACM workshop on Performance monitoring and measurement of heterogeneous wireless and wired* 

Kaabi, F.; Ghannay, S. & Filali F. (2010). Channel Allocation and Routing in Wireless Mesh Networks: A survey and qualitative comparison between schemes, In: *International* 

Pollak, S. & Wieser, V. (2012). Interference Reduction Channel Assignment Algorithm for Multi-Interface Wireless Mesh Networks, In: *Proceedings of 22nd International Conference* 

Prodan, A. & Mirchandani, V. (2009). Channel Assignment Techniques for 802.11 based Multi-Radio Wireless Mesh Networks, in *Handbook of Wireless Mesh Networks*, Springer

Ramachandran, K. N.; Belding, E. M.; Almeroth, K. C. & Buddhikot, M. M. (2006). Interference-Aware Channel Assignment in Multi-Radio Wireless Mesh Networks, INFOCOM 2006. *25th IEEE International Conference on Computer Communications.* 

Raniwala, A.; Gopalan, K. & Chiueh T. (2004). Centralized Channel Assignment and Routing Algorithms for Multi-Channel Wireless Mesh Networks, In Acm *Sigmobile* 

Skalli, H.; Das, S. K.; Lenzini L. & Conti M. (2006). Traffic and interference aware channel

So, J. & Vaidya, N. H. (2004). Multi-channel MAC for Ad Hoc Networks: Handling Multichannel Hidden Terminals using a Single Transceiver, In *ACM Mobihoc*, 2004 Wei Yahuan; Taoshen Li & Zhihui Ge (2011). A Channel Assignment Algorithm for Wireless Mesh Networks Using the Maximum Flow Approach. In: *Journal of networks*, Vol. 6, No.

Yulong Chen; Ning Xie; Gongbin Qian & Hui Wang (2010). Channel assignment schemes in Wireless Mesh Networks, In: *Mobile Congress (GMC)*, 2010 Global , vol., no., pp.1-5, 18-

assignment for multi-radio Wireless Mesh Networks, Technical report

*Journal of Wireless & Mobile Networks (IJWMN)*, Vol. 2, No. 1, pp. 132-150, 2010 Marina, M. K. & Das, S. R. (2005). A topology control approach for utilizing multiple channels in multi-radio wireless mesh networks, *Broadband Networks Vol.1*, BroadNets

*networks*, p.120-127, October 26-26, 2009, Tenerife, Canary Islands, Spain

ns-2 (2008), The Network Simulator ns-2, http://www.isi.edu/nsnam/ns/

*"Radioelektronika 2012"*, Brno, Czech republic (paper accepted)

Publishers, ISBN 978-1-84800-908-0, London 2009

The concept of wireless mesh networks (WMN) has emerged as a promising technology for the provision of affordable and low-cost solutions for a wide range of applications such as broadband wireless internet access in developing regions with no or limited wired infrastructure, security surveillance, and emergency networking, One concrete example is WMNs for public safety teams like firefighters who can still be connected with the help of mesh nodes mounted on street poles even if all infrastructure communications fail. The main reason for this vast acceptance of mesh networks in the industry and academia is because of their self-maintenance feature and the low cost of wireless routers. In addition, the self-forming features of WMNs make the deployment of a mesh network easy thereby enabling large-scale networks. Mesh networks which are of most commercial interests are characterized as fixed backbone WMNs where mesh nodes (routers or access points) are generally static and are mostly supplied by a permanent power source. Such a wireless mesh network architecture is illustrated in Figure 1, consisting of mesh routers, clients, and gateway nodes. Mesh routers (MR) communicate with peers in a multi hop fashion such that packets are mostly transmitted over multiple wireless links (hops). Therefore, nodes forward packets to other nodes that are on the route but may not be within direct transmission range of each other. Routers which are connected to the outside world are called gateway nodes (GWN). These GWNs carry traffic in and out of the mesh network. The collection of such routers and gateway nodes connected together in a multi hop fashion form the basis for an infrastructure WMN (also called backbone mesh). Moreover, the multi hop packet transmission in an infrastructure WMN extends the area of wireless broadband coverage without wiring the network; thus WMNs can be used as extensions to cellular networks, ad hoc networks (MANET), sensor and vehicular networks, IEEE 802.11 WLANs (Wi-Fi), and IEEE 802.16 based broadband wireless (WiMax) networks [1].

© 2012 Bokhari and Záruba, licensee InTech. This is an open access chapter distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. © 2012 The Author(s). Licensee InTech. This chapter is distributed under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Partially Overlapping Channel Assignments in Wireless Mesh Networks 105

(OCs) to radios belonging to different nodes, thus improving network capacity. However, such networks severely suffer from network disconnections due to having a single radio at each node possibly configured at different channels. In, MRMC-WMNs, with the availability of off-the-shelf, low cost, IEEE 802.11 based networking hardware, it is possible to incorporate multiple radio interfaces operating on different radio channels on a single mesh router. This enables a potentially large improvement in the capacity of the WMN (compared to all

Wireless mesh networks, particularly infrastructure WMNs, have some unique characteristics that set them apart from other wireless networks, such as MANETs and sensor networks. For example, nodes (at least relay nodes) in a typical infrastructure mesh network are generally static and have no significant constraints on power consumption, as opposed to MANETs, where nodes have limited energy and are mostly mobile. Similarly, due to the shared nature of the wireless medium, nodes compete with each other for channel access when they transmit on the same channel resulting in possible interference among the nodes. Unlike MA-NETs, where the general traffic model describes traffic flows between any pair of mobile nodes, in WMNs data flows are typically between mesh nodes and GWNs. In general, in WMNs certain paths and nodes are much more likely to be saturated as the distribution of flows over nodes is less uniform compared to that in MANETs. Therefore, load balancing is

In a typical multi radio mesh network, the total number of radios within the network is usually significantly higher than the number of available channels in the network (e.g. only 11 channels are available in the U.S.A. for IEEE 802.11b/g). This forces many links to operate on the same (set of) channels, resulting in possible interference among transmissions. The existence of such interference if not accounted for, can affect the capacity of the network. Therefore, understanding and mitigating interference has become one of the fundamental issues in WMNs; recently a number of channel assignment (CA) solutions have been pro-

The problem of channel assignment (frequency assignment) has been widely studied in cellular networks [2]. However, with the proliferation of IEEE 802.11 based technologies in the wireless arena (WLANs, sensor networks, WMNs), the need for channel assignment solutions outside of cellular networks has surfaced. CA algorithms are usually designed based on the peculiar characteristics of individual networks; since the differences in characteristics are vast, CA algorithms for WMNs must be significantly different from those of cellular networks. For example, base stations in a cellular network are typically connected by cables, whereas mesh nodes in a mesh network are connected wirelessly (and usually on the same channels as are used for providing service). This brings up several interference issues in mesh networks between mesh nodes which are not found in cellular networks between base stations (as in cellular networks BSs are not competing for the shared medium as they have dedicated bandwidth for intra-BS communication). The bottleneck in cellular networks is from the base stations to the client devices, whereas, in WMNs, the bottleneck is usually inside the mesh backbone, typically along the route from the mesh routers to the gateway

of utmost importance to avoid hot spots and to increase network utilization.

posed to address this problem [5, 10-13, 15-20, 33-35].

the previous forms of mesh networks) [20].

**Figure 1.** A typical wireless mesh network architecture.

WMNs can be classified based on the number of radios on each mesh router. In single-radio mesh networks, each node is equipped with only one radio. In multi-radio mesh networks, multiple radios are installed on each mesh node in the backbone mesh. Depending upon the radio to channel configuration (also called interface to channel assignment), mesh networks can be further classified into single-radio single-channel (SRSC), single-radio multi-channel (SRMC), and multi-radio multi-channel (MRMC) wireless mesh networks. (Note, that we did not list multi-radio single-channel WMNs as that would mean that nodes are equipped with multiple radios but all of the radios in the network are configured on the same single channel defeating any purpose of multi-radios.) In a SRSC-WMN, as the name suggests, all nodes are configured to use the same wireless channel. This ensures network connectivity; however, capacity of the network is greatly affected as all nodes are competing to access the same channel. Therefore, interference minimization is the major issue in such networks. SRMC-WMNs can achieve parallel transmissions by assigning different orthogonal channels (OCs) to radios belonging to different nodes, thus improving network capacity. However, such networks severely suffer from network disconnections due to having a single radio at each node possibly configured at different channels. In, MRMC-WMNs, with the availability of off-the-shelf, low cost, IEEE 802.11 based networking hardware, it is possible to incorporate multiple radio interfaces operating on different radio channels on a single mesh router. This enables a potentially large improvement in the capacity of the WMN (compared to all the previous forms of mesh networks) [20].

104 Wireless Mesh Networks – Efficient Link Scheduling, Channel Assignment and Network Planning Strategies

**Figure 1.** A typical wireless mesh network architecture.

WMNs can be classified based on the number of radios on each mesh router. In single-radio mesh networks, each node is equipped with only one radio. In multi-radio mesh networks, multiple radios are installed on each mesh node in the backbone mesh. Depending upon the radio to channel configuration (also called interface to channel assignment), mesh networks can be further classified into single-radio single-channel (SRSC), single-radio multi-channel (SRMC), and multi-radio multi-channel (MRMC) wireless mesh networks. (Note, that we did not list multi-radio single-channel WMNs as that would mean that nodes are equipped with multiple radios but all of the radios in the network are configured on the same single channel defeating any purpose of multi-radios.) In a SRSC-WMN, as the name suggests, all nodes are configured to use the same wireless channel. This ensures network connectivity; however, capacity of the network is greatly affected as all nodes are competing to access the same channel. Therefore, interference minimization is the major issue in such networks. SRMC-WMNs can achieve parallel transmissions by assigning different orthogonal channels Wireless mesh networks, particularly infrastructure WMNs, have some unique characteristics that set them apart from other wireless networks, such as MANETs and sensor networks. For example, nodes (at least relay nodes) in a typical infrastructure mesh network are generally static and have no significant constraints on power consumption, as opposed to MANETs, where nodes have limited energy and are mostly mobile. Similarly, due to the shared nature of the wireless medium, nodes compete with each other for channel access when they transmit on the same channel resulting in possible interference among the nodes. Unlike MA-NETs, where the general traffic model describes traffic flows between any pair of mobile nodes, in WMNs data flows are typically between mesh nodes and GWNs. In general, in WMNs certain paths and nodes are much more likely to be saturated as the distribution of flows over nodes is less uniform compared to that in MANETs. Therefore, load balancing is of utmost importance to avoid hot spots and to increase network utilization.

In a typical multi radio mesh network, the total number of radios within the network is usually significantly higher than the number of available channels in the network (e.g. only 11 channels are available in the U.S.A. for IEEE 802.11b/g). This forces many links to operate on the same (set of) channels, resulting in possible interference among transmissions. The existence of such interference if not accounted for, can affect the capacity of the network. Therefore, understanding and mitigating interference has become one of the fundamental issues in WMNs; recently a number of channel assignment (CA) solutions have been proposed to address this problem [5, 10-13, 15-20, 33-35].

The problem of channel assignment (frequency assignment) has been widely studied in cellular networks [2]. However, with the proliferation of IEEE 802.11 based technologies in the wireless arena (WLANs, sensor networks, WMNs), the need for channel assignment solutions outside of cellular networks has surfaced. CA algorithms are usually designed based on the peculiar characteristics of individual networks; since the differences in characteristics are vast, CA algorithms for WMNs must be significantly different from those of cellular networks. For example, base stations in a cellular network are typically connected by cables, whereas mesh nodes in a mesh network are connected wirelessly (and usually on the same channels as are used for providing service). This brings up several interference issues in mesh networks between mesh nodes which are not found in cellular networks between base stations (as in cellular networks BSs are not competing for the shared medium as they have dedicated bandwidth for intra-BS communication). The bottleneck in cellular networks is from the base stations to the client devices, whereas, in WMNs, the bottleneck is usually inside the mesh backbone, typically along the route from the mesh routers to the gateway

Partially Overlapping Channel Assignments in Wireless Mesh Networks 107

carefully coordinated; the key issue lies in the fact that the interference between adjacent channels has to be considered. This needs to be done intelligently so that channel capacity is maximized, otherwise the shared nature of wireless medium can lead to serious performance degradation of the whole mesh network. Thus, recently POCs for channel

Within the scope of this chapter, we focus on the problem of channel assignment using partially overlapping channels in the context of both single- and multi-radio WMNs. The rest of the chapter is organized as follows. Section 2 describes different types of interferences that may exist in a typical WMN. Section 3 demonstrates the benefits of using partially overlapping channels for channel assignment in WMNs with the help of experiments performed on a real testbed. In Section 4, we provide a comprehensive review of some of the recent wellknown channel assignment schemes exploiting POCs in WMNs and classify these POCbased CA schemes according to their most prominent attributes together with the objectives and limitations of each of the approaches. In Section 5, we discuss open issues and challenges in the design of partially overlapping channel assignment schemes, followed by the chap-

assignment in wireless networks has received some attention [5, 10-13, 15-20].

**Figure 2.** IEEE 802.11b/g channels, showing the three orthogonal channels in bold

In a typical WMN, flows on links belonging to different nodes compete with each other to access the wireless medium. This results in possible interference among the nodes therefore severely affecting network performance. Multiple types of interferences exist in WMNs depending on flow characteristics and on interface to channel configurations. We first explain what the different types of flow interferences are particularly in infrastructure WMNs. We will also present another interference classification in mesh networks based on the configura-

This type of interference occurs when neighboring nodes carrying different flows compete for channel access when they transmit on the same channel as depicted in Figure 3(a). This

tion of the channels to radios and also on the number of radios installed in nodes.

**2. Interference in Wireless Networks** 

**2.1. Flow based interference** 

*2.1.1. Inter-flow Interference* 

ter's conclusion in Section 6.

nodes. In addition nearby BSs are usually configured on completely orthogonal channels (OCs) to avoid interference; this is rarely possible in backbone meshes, as the nodal density of a typical WMN can be high and the number of available orthogonal channels is limited. Most existing deployed mesh networks are IEEE 802.11 technology based; among the standards of IEEE 802.11, the most widely used are IEEE 802.11b/g, which support up to 14 channels in the unlicensed Industrial, Scientific, and Medical (ISM) radio bands at the nominal 2.4 GHz carrier frequency [32]. Out of these 14 channels, only 11 channels are available for use in the U.S.A., 13 channels are open in EU, while Japan has made all of them available. Figure 2. shows the 2.4 GHz ISM band's division into 11 IEEE 802.11b/g channels in the U.S.A.; the channel numbers have a one-to-one relationship with the corresponding center frequency of that channel. (For example, channel 6 operates at 2.437GHz.) Each channel's bandwidth is 22 MHz and each channel's center frequency is separated from the next channel's by 5MHz. Therefore, in general, a channel overlaps with 4 of its neighboring channels leaving only three non-overlapping (orthogonal) channels, i.e., channels 1, 6, and 11 as depicted in Figure 2. Similarly, IEEE 802.11a operates in 5GHz ISM band and provides 12 orthogonal channels, but since it operates in a higher frequency band, it has a shorter range as opposed to 802.11b/g (higher frequencies in general have higher inabilities to penetrate walls and obstructions). Recently, the IEEE 802.11n standard was proposed which operates in both the 2.4GHz and 5 GHz bands and provides legacy support to devices operating based on previous standards (b/g). It provides data rates of up to 600Mbps using Multiple Input, Multiple Output (MIMO) technology with Orthogonal Frequency Division Multiplexing (OFDM).

Most existing research on CA algorithms in WMNs has been focused on assigning orthogonal (non-overlapping) channels [33-35] to links belonging to neighboring nodes in order to minimize the interference in the network. Since, links operating on orthogonal channels do not interfere at all, multiple parallel transmissions can be possible resulting in overall network throughput improvement. The number of non-overlapping channels in commodity wireless platforms such as 802.11b/g is very small (again, only three orthogonal channels out of total 11 channels) while nodal density in a typical MRMC-WMN is high. This realization has recently drawn significant attention to the study of partially overlapped channels (POC) for channel assignment [5]. The basic idea is to make all channels available to nodes for channel selection as a result of which, partially overlapped channels may be employed. This could enable multiple concurrent transmissions on radios configured on POCs and therefore could increase network capacity assuming that the interference is lessened in POCs compared to completely overlapping channels.

Previously, an algorithm for channel assignment based solely on orthogonal channels had to deal with only co-channel interference. However, one of the major issues in designing efficient channel assignment schemes using POCs is the adjacent channel interference, which is the interference between two neighbors configured on adjacent (partially overlapping) channels. The effect of such adjacent channel interference has a direct relationship with the geographical location of these two nodes, i.e., the farther two nodes are apart, the less interference is created on adjacent channels. Nonetheless, the assignment of orthogonal and non-orthogonal channels in high density mesh networks needs to be carefully coordinated; the key issue lies in the fact that the interference between adjacent channels has to be considered. This needs to be done intelligently so that channel capacity is maximized, otherwise the shared nature of wireless medium can lead to serious performance degradation of the whole mesh network. Thus, recently POCs for channel assignment in wireless networks has received some attention [5, 10-13, 15-20].

Within the scope of this chapter, we focus on the problem of channel assignment using partially overlapping channels in the context of both single- and multi-radio WMNs. The rest of the chapter is organized as follows. Section 2 describes different types of interferences that may exist in a typical WMN. Section 3 demonstrates the benefits of using partially overlapping channels for channel assignment in WMNs with the help of experiments performed on a real testbed. In Section 4, we provide a comprehensive review of some of the recent wellknown channel assignment schemes exploiting POCs in WMNs and classify these POCbased CA schemes according to their most prominent attributes together with the objectives and limitations of each of the approaches. In Section 5, we discuss open issues and challenges in the design of partially overlapping channel assignment schemes, followed by the chapter's conclusion in Section 6.

**Figure 2.** IEEE 802.11b/g channels, showing the three orthogonal channels in bold

## **2. Interference in Wireless Networks**

In a typical WMN, flows on links belonging to different nodes compete with each other to access the wireless medium. This results in possible interference among the nodes therefore severely affecting network performance. Multiple types of interferences exist in WMNs depending on flow characteristics and on interface to channel configurations. We first explain what the different types of flow interferences are particularly in infrastructure WMNs. We will also present another interference classification in mesh networks based on the configuration of the channels to radios and also on the number of radios installed in nodes.

#### **2.1. Flow based interference**

#### *2.1.1. Inter-flow Interference*

106 Wireless Mesh Networks – Efficient Link Scheduling, Channel Assignment and Network Planning Strategies

technology with Orthogonal Frequency Division Multiplexing (OFDM).

lessened in POCs compared to completely overlapping channels.

Most existing research on CA algorithms in WMNs has been focused on assigning orthogonal (non-overlapping) channels [33-35] to links belonging to neighboring nodes in order to minimize the interference in the network. Since, links operating on orthogonal channels do not interfere at all, multiple parallel transmissions can be possible resulting in overall network throughput improvement. The number of non-overlapping channels in commodity wireless platforms such as 802.11b/g is very small (again, only three orthogonal channels out of total 11 channels) while nodal density in a typical MRMC-WMN is high. This realization has recently drawn significant attention to the study of partially overlapped channels (POC) for channel assignment [5]. The basic idea is to make all channels available to nodes for channel selection as a result of which, partially overlapped channels may be employed. This could enable multiple concurrent transmissions on radios configured on POCs and therefore could increase network capacity assuming that the interference is

Previously, an algorithm for channel assignment based solely on orthogonal channels had to deal with only co-channel interference. However, one of the major issues in designing efficient channel assignment schemes using POCs is the adjacent channel interference, which is the interference between two neighbors configured on adjacent (partially overlapping) channels. The effect of such adjacent channel interference has a direct relationship with the geographical location of these two nodes, i.e., the farther two nodes are apart, the less interference is created on adjacent channels. Nonetheless, the assignment of orthogonal and non-orthogonal channels in high density mesh networks needs to be

nodes. In addition nearby BSs are usually configured on completely orthogonal channels (OCs) to avoid interference; this is rarely possible in backbone meshes, as the nodal density of a typical WMN can be high and the number of available orthogonal channels is limited. Most existing deployed mesh networks are IEEE 802.11 technology based; among the standards of IEEE 802.11, the most widely used are IEEE 802.11b/g, which support up to 14 channels in the unlicensed Industrial, Scientific, and Medical (ISM) radio bands at the nominal 2.4 GHz carrier frequency [32]. Out of these 14 channels, only 11 channels are available for use in the U.S.A., 13 channels are open in EU, while Japan has made all of them available. Figure 2. shows the 2.4 GHz ISM band's division into 11 IEEE 802.11b/g channels in the U.S.A.; the channel numbers have a one-to-one relationship with the corresponding center frequency of that channel. (For example, channel 6 operates at 2.437GHz.) Each channel's bandwidth is 22 MHz and each channel's center frequency is separated from the next channel's by 5MHz. Therefore, in general, a channel overlaps with 4 of its neighboring channels leaving only three non-overlapping (orthogonal) channels, i.e., channels 1, 6, and 11 as depicted in Figure 2. Similarly, IEEE 802.11a operates in 5GHz ISM band and provides 12 orthogonal channels, but since it operates in a higher frequency band, it has a shorter range as opposed to 802.11b/g (higher frequencies in general have higher inabilities to penetrate walls and obstructions). Recently, the IEEE 802.11n standard was proposed which operates in both the 2.4GHz and 5 GHz bands and provides legacy support to devices operating based on previous standards (b/g). It provides data rates of up to 600Mbps using Multiple Input, Multiple Output (MIMO)

> This type of interference occurs when neighboring nodes carrying different flows compete for channel access when they transmit on the same channel as depicted in Figure 3(a). This

effectively means that whenever a node is involved in a transmission; its neighboring nodes should not communicate at the same time.

Partially Overlapping Channel Assignments in Wireless Mesh Networks 109

Self-interference is defined as a transmission from a node interfering with one of its own transmissions. This is related to situations when nodes are equipped with multiple radios in a mesh network. Parallel communication cannot be achieved using multiple radios installed on a node, unless they are configured on **completely orthogonal** channels as shown in Figure 4(c).

All of the above types of interferences have to be considered when designing channel assignment algorithms to exploit the full potential of the available wireless spectrum. Therefore, the first step in developing mechanisms which take advantage of the partial overlap is to build a model that captures the channel overlap in a quantitative fashion.

**Figure 4.** Types of interferences (a) co-channel interference (b) adjacent channel interference (c) self-

In this section, we will discuss the benefits of using POCs in WMNs. First, we will explain what the different scenarios are, where the use of partial overlap among channels will be useful. We follow that by a quick testbed experiment to demonstrate the effectiveness of

Mishra, et al., in [6] have performed detailed experiments to demonstrate the effectiveness of using partial overlap among channels in WMNs. The authors have measured the signal to noise ratio (SNR) of two communicating nodes configured on adjacent channels and mapped them onto a normalized [0,1] scale with 0 representing the minimum signal received. Their

Channel 1 2 3 4 5 6 7 8 9 10 11 Normalized SNR (I-factor) 0 0.22 0.60 0.72 0.77 1.0 0.96 0.77 0.66 0.39 0

A typical bandwidth of an IEEE 802.11b channel which uses direct sequence spread spectrum (DSSS) is 44MHz. It is distributed equally on each side of the center frequency of that

**Table 1.** SNR of transmission made on channel 6 as received on channels 1 ... 11.

**3. Benefits of using Partially Overlapped Channels** 

*2.2.3. Self-Interference (SI)* 

interference

using POCs in WMNs.

results are shown in Table I.

#### *2.1.2. Intra-flow Interference*

Nodes on the path of a same flow compete with each other for channel access when they transmit on the same channel. This is referred to as intra-flow interference and is shown in Figure 3(b).

**Figure 3.** Flow based interference. (a) Inter-flow interference (b) Intra-flow interference

#### **2.2. Interference based on interface to channel configuration**

A wireless mesh network utilizing both orthogonal and non-orthogonal channels may suffer from interferences which can be characterized as follows.

#### *2.2.1. Co-channel Interference (CCI)*

Co-channel interference is the most common type of interference that exists in almost all wireless networks (depicted in Figure 4-a). It refers to the fact that radios belonging to two nodes, operating on the same channel would interfere with each other, if they are within the interference range of each other. This effectively means that parallel communications from two separate in-range nodes is not possible.

#### *2.2.2. Adjacent Channel Interference (ACI)*

We talk about adjacent channel interference when radios on two nearby nodes are configured to partially overlapping channels. For example, in Figure 4(b), a radio on node *A* is configured on channel-4 while another radio at neighboring node *C* is configured on channel-1; then the transmission from either node would experience some sort of partial interference. This type of interference also restricts parallel communication depending upon the channel separation and the physical distance between the two nodes.

#### *2.2.3. Self-Interference (SI)*

108 Wireless Mesh Networks – Efficient Link Scheduling, Channel Assignment and Network Planning Strategies

**Figure 3.** Flow based interference. (a) Inter-flow interference (b) Intra-flow interference

A wireless mesh network utilizing both orthogonal and non-orthogonal channels may suffer

Co-channel interference is the most common type of interference that exists in almost all wireless networks (depicted in Figure 4-a). It refers to the fact that radios belonging to two nodes, operating on the same channel would interfere with each other, if they are within the interference range of each other. This effectively means that parallel communications from

We talk about adjacent channel interference when radios on two nearby nodes are configured to partially overlapping channels. For example, in Figure 4(b), a radio on node *A* is configured on channel-4 while another radio at neighboring node *C* is configured on channel-1; then the transmission from either node would experience some sort of partial interference. This type of interference also restricts parallel communication depending upon the

**2.2. Interference based on interface to channel configuration** 

channel separation and the physical distance between the two nodes.

from interferences which can be characterized as follows.

*2.2.1. Co-channel Interference (CCI)* 

two separate in-range nodes is not possible.

*2.2.2. Adjacent Channel Interference (ACI)* 

should not communicate at the same time.

*2.1.2. Intra-flow Interference* 

Figure 3(b).

effectively means that whenever a node is involved in a transmission; its neighboring nodes

Nodes on the path of a same flow compete with each other for channel access when they transmit on the same channel. This is referred to as intra-flow interference and is shown in Self-interference is defined as a transmission from a node interfering with one of its own transmissions. This is related to situations when nodes are equipped with multiple radios in a mesh network. Parallel communication cannot be achieved using multiple radios installed on a node, unless they are configured on **completely orthogonal** channels as shown in Figure 4(c).

All of the above types of interferences have to be considered when designing channel assignment algorithms to exploit the full potential of the available wireless spectrum. Therefore, the first step in developing mechanisms which take advantage of the partial overlap is to build a model that captures the channel overlap in a quantitative fashion.

**Figure 4.** Types of interferences (a) co-channel interference (b) adjacent channel interference (c) selfinterference

### **3. Benefits of using Partially Overlapped Channels**

In this section, we will discuss the benefits of using POCs in WMNs. First, we will explain what the different scenarios are, where the use of partial overlap among channels will be useful. We follow that by a quick testbed experiment to demonstrate the effectiveness of using POCs in WMNs.

Mishra, et al., in [6] have performed detailed experiments to demonstrate the effectiveness of using partial overlap among channels in WMNs. The authors have measured the signal to noise ratio (SNR) of two communicating nodes configured on adjacent channels and mapped them onto a normalized [0,1] scale with 0 representing the minimum signal received. Their results are shown in Table I.


**Table 1.** SNR of transmission made on channel 6 as received on channels 1 ... 11.

A typical bandwidth of an IEEE 802.11b channel which uses direct sequence spread spectrum (DSSS) is 44MHz. It is distributed equally on each side of the center frequency of that channel i.e. 22MHz on each side. A transmit spectrum mask (band pass filter) is applied to the signal at the transmitting station (with a typical example shown in Figure 5) which is basically used by the transmitter to limit the output power on nearby frequencies. As it can be seen in the figure, the mask is set to 0dB at the center frequency where signals are passed without any attenuation. However, at frequencies beyond 11MHz on either side of the center frequency, the signal's power is attenuated by as much as 30dB and at 22MHz as much as 50dB. The receiver also uses a band pass filter centered around the nominal transmission frequency of the channel. Three scenarios are discussed in [6] where the use of partial overlap among the channels can be useful in the context of wireless mesh networks:

Partially Overlapping Channel Assignments in Wireless Mesh Networks 111

configured to overlapping channels in order to avoid network disconnection and also to

Next we will show results from experiments performed on a real testbed in order to evaluate the benefits of using partially overlapping channels in mesh networks. Our experimental testbed consists of four Linksys WRT54GLv1.1 wireless routers, each equipped with one radio. We installed the Freifunk firmware [28] on these routers for more freedom in our experiments. We created two point-to-point networks between two router pairs and thus formed two links each consisting of two routers as shown in Figure 6. Link-1 belongs to Pair-1 and Link-2 belongs to Pair-2. Each radio on Link-1 is fixed on channel 6; we varied the channels of Pair-2 from 1 to 6. The distance between nodes belonging to the same link is kept constant throughout the experiment. Pair-1 nodes have fixed locations while Pair-2 is moved to various distances from Pair-1 ranging from 5 to 30 meters (but Pair-2 nodes are kept equidistance to each other during the experiments). UDP and TCP traffic is generated on both links lasting for 10 seconds. The throughput on Link-2 is measured and the results are averaged over several runs. Three different IEEE 802.11b defined data rates are used for

Figure 7(a), (b), and (c) show the UDP throughput on Link-2 with different channel separations for the three data rates. It can be seen that as the distance between the two interfering links is increased, the throughput increases due to the reduced amount of interference. In this setup we did not see any further improvements when nodes were more than 30 meters apart. However, the same maximum throughput can be achieved at significantly lower distances with increased channel separation between the two links. For example, at about 20 meters, Link-2 achieves the maximum benchmark throughput, when the channel separation between the two links is three. For data rates 5.5Mbps and 11Mbps, we notice similar results; however, maximum throughput can be achieved by eliminating interference at a much lower distances i.e. about 15 meters, when the channel separation is three as compared to 30 meters, when both the channels are separated by only one. Figures 8 (a), (b), and (c) show the comparable results

avoid any channel switching overhead.

conducting the experiments, i.e. 2Mbps, 5.5Mbps and 11Mbps.

when TCP traffic is used on all the three 802.11b data rates.

**3.1. Experimental evaluation** 

**Figure 6.** POC measurement testbed


**Figure 5.** A typical IEEE 802.11b transmit spectrum mask

Later, in [3], the same authors have shown the advantage of using POCs in two different types of networks, i.e., WLANs and WMNs. In a WLAN setup, nearby access points can be assigned POCs such that the signal attenuation due to the overlap degrades to a tolerable level. In other words, the interference range of APs is reduced as perceived by neighboring APs operating on a partially overlapping channel. This provides efficient spatial re-use of channels and more APs can operate concurrently providing better service to clients. Similarly, in a single radio WMN environment, throughput can be improved when nodes can be configured to overlapping channels in order to avoid network disconnection and also to avoid any channel switching overhead.

#### **3.1. Experimental evaluation**

110 Wireless Mesh Networks – Efficient Link Scheduling, Channel Assignment and Network Planning Strategies

lap among the channels can be useful in the context of wireless mesh networks:

provements.

in higher peak throughputs.

**Figure 5.** A typical IEEE 802.11b transmit spectrum mask

channel i.e. 22MHz on each side. A transmit spectrum mask (band pass filter) is applied to the signal at the transmitting station (with a typical example shown in Figure 5) which is basically used by the transmitter to limit the output power on nearby frequencies. As it can be seen in the figure, the mask is set to 0dB at the center frequency where signals are passed without any attenuation. However, at frequencies beyond 11MHz on either side of the center frequency, the signal's power is attenuated by as much as 30dB and at 22MHz as much as 50dB. The receiver also uses a band pass filter centered around the nominal transmission frequency of the channel. Three scenarios are discussed in [6] where the use of partial over-

 *Multi-channel communication:* The first scenario is when a node can communicate with two of its neighboring nodes configured on orthogonal channels (OCs) by operating on a partial overlapping channel. Basically, for a little reduction in throughput, one can use partially overlapping channels and this can give flexibility in topology construction while reducing the extra overhead in channel switching to enable communication. *Throughput improvement:* The second scenario is when nodes in a mesh network have only one radio and therefore, they can be configured to only one channel at a time. There is a possibility of network disconnection while assigning different channels to nodes in the network. Channels with partial overlap can be assigned to nodes in such a manner that improves the overall network throughput capacity. In this way, the assignment of partially overlapping channels has to be intelligent enough to utilize the maximum bandwidth available and therefore can result in significant throughput im-

 *Channel re-use:* Shorter ranges for frequency reuse can be obtained if two interfering links are assigned partially overlapping channels rather than orthogonal channels. It is possible to significantly improve the overall channel re-use (i.e., by reducing the distance between nodes using POCs) by careful assignment of channels which will result

Later, in [3], the same authors have shown the advantage of using POCs in two different types of networks, i.e., WLANs and WMNs. In a WLAN setup, nearby access points can be assigned POCs such that the signal attenuation due to the overlap degrades to a tolerable level. In other words, the interference range of APs is reduced as perceived by neighboring APs operating on a partially overlapping channel. This provides efficient spatial re-use of channels and more APs can operate concurrently providing better service to clients. Similarly, in a single radio WMN environment, throughput can be improved when nodes can be Next we will show results from experiments performed on a real testbed in order to evaluate the benefits of using partially overlapping channels in mesh networks. Our experimental testbed consists of four Linksys WRT54GLv1.1 wireless routers, each equipped with one radio. We installed the Freifunk firmware [28] on these routers for more freedom in our experiments. We created two point-to-point networks between two router pairs and thus formed two links each consisting of two routers as shown in Figure 6. Link-1 belongs to Pair-1 and Link-2 belongs to Pair-2. Each radio on Link-1 is fixed on channel 6; we varied the channels of Pair-2 from 1 to 6. The distance between nodes belonging to the same link is kept constant throughout the experiment. Pair-1 nodes have fixed locations while Pair-2 is moved to various distances from Pair-1 ranging from 5 to 30 meters (but Pair-2 nodes are kept equidistance to each other during the experiments). UDP and TCP traffic is generated on both links lasting for 10 seconds. The throughput on Link-2 is measured and the results are averaged over several runs. Three different IEEE 802.11b defined data rates are used for conducting the experiments, i.e. 2Mbps, 5.5Mbps and 11Mbps.

**Figure 6.** POC measurement testbed

Figure 7(a), (b), and (c) show the UDP throughput on Link-2 with different channel separations for the three data rates. It can be seen that as the distance between the two interfering links is increased, the throughput increases due to the reduced amount of interference. In this setup we did not see any further improvements when nodes were more than 30 meters apart. However, the same maximum throughput can be achieved at significantly lower distances with increased channel separation between the two links. For example, at about 20 meters, Link-2 achieves the maximum benchmark throughput, when the channel separation between the two links is three. For data rates 5.5Mbps and 11Mbps, we notice similar results; however, maximum throughput can be achieved by eliminating interference at a much lower distances i.e. about 15 meters, when the channel separation is three as compared to 30 meters, when both the channels are separated by only one. Figures 8 (a), (b), and (c) show the comparable results when TCP traffic is used on all the three 802.11b data rates.

Partially Overlapping Channel Assignments in Wireless Mesh Networks 113

**Figure 8.** TCP throughput of two interfering links as a function of channel separation. (a) 2 Mbps (b) 5.5

Mbps (c) 11 Mbps

**Figure 9.** Interference range as a function of data rates

From these results, we can extrapolate the interference ranges of nodes with varying channel separations and at different data rates; this comprehension is shown in Figure 9. Each point in the graph represents the minimum distance that is required between the two links in order for them to experience no interference and achieve maximum throughput when they are on particular partially overlapping channels (with a given channel separation). We can observe that the interference ranges are decreasing with increasing channel separation and increasing data rates. From these measurements, we can empirically conclude that the interference range of nodes operating on POCs is significantly less than the range when they are on the same channel. (Similar experiments have been performed before in [3, 5-7, 16]; however, those experiments were done either on wireless card equipped computers or a computer attached to an access point. We believe, that our setup is easier to reproduce and is more representative for a WMN and thus provides a better understanding of POCs in mesh networks.)

Therefore, there is a tradeoff between efficient utilization of the wireless spectrum and a slight decrease in the throughput. An intelligent assignment of partially overlapping channels can decrease the impact of interference, eventually resulting in more efficient utilization of the spectrum.

**Figure 7.** UDP throughput of two interfering links as a function of channel separation. (a) 2 Mbps (b) 5.5 Mbps (c) 11 Mbps

**Figure 8.** TCP throughput of two interfering links as a function of channel separation. (a) 2 Mbps (b) 5.5 Mbps (c) 11 Mbps

**Figure 9.** Interference range as a function of data rates

understanding of POCs in mesh networks.)

of the spectrum.

5.5 Mbps (c) 11 Mbps

From these results, we can extrapolate the interference ranges of nodes with varying channel separations and at different data rates; this comprehension is shown in Figure 9. Each point in the graph represents the minimum distance that is required between the two links in order for them to experience no interference and achieve maximum throughput when they are on particular partially overlapping channels (with a given channel separation). We can observe that the interference ranges are decreasing with increasing channel separation and increasing data rates. From these measurements, we can empirically conclude that the interference range of nodes operating on POCs is significantly less than the range when they are on the same channel. (Similar experiments have been performed before in [3, 5-7, 16]; however, those experiments were done either on wireless card equipped computers or a computer attached to an access point. We believe, that our setup is easier to reproduce and is more representative for a WMN and thus provides a better

Therefore, there is a tradeoff between efficient utilization of the wireless spectrum and a slight decrease in the throughput. An intelligent assignment of partially overlapping channels can decrease the impact of interference, eventually resulting in more efficient utilization

**Figure 7.** UDP throughput of two interfering links as a function of channel separation. (a) 2 Mbps (b)
