**2. Related works**

A major problem facing multi-hop wireless networks is the interference between adjacent links. The throughput of a single-radio single-channel wireless network has been studied in [37]. The authors formalized it as a multi-commodity flow problem with constraints from conflict graph, which is NP hard, and gave an upper bound and a lower bound of the problem.

There have been many studies on how to assign limited channels to network interfaces in a multi-radio multi-channel wireless mesh network as to minimize interference and maximize throughput. They differ in several assumptions made in WMNs, and therefore in the models and related solutions.

One approach assumes a known traffic profile in the network, because the aggregate traffic load of each mesh router changes infrequently. The authors of [26] proposed an iterative approach to solve the joint routing and channel assignment problem. Heuristic techniques are used to estimate the traffic load in each link. The algorithm starts with an initial estimation of the expected traffic load and iterates over both channel assignment and routing until the bandwidth allocated to each virtual link matches its expected load. While this scheme presents a method for channel allocation that incorporates connectivity and traffic patterns, the assignment of channels on links may cause a ripple effect whereby already assigned links have to be revisited, thus, increasing the time complexity of the scheme. Moreover, this approach is performed during the network plan and assumes that the traffic profile is known. The centralized flow-based and rate channel assignment algorithm is proposed in a paper by [3]. The agreed heuristic algorithm is used for channel assignment rate. [6] enhances the [26] centralized algorithm to support automatic and fast failure recovery. The failure recovery mechanism is located at the gateway and all nodes send periodic messages to the gateway. In the case where the gateway does not receive a message during a period of time from node *x*, it deletes the corresponding information, node id, position, rate, and then runs the algorithm to update the gateway tables. Based on the new tables, it recalculates the link ranking and channel assignment. However, in general, the centralized approach causes a high computation overhead at the centric node and it is unwieldy in use due to the need for gathering network information. Moreover, most of them are static assignment which is not optimally utilizing the limited number of available non-overlapping channels. In contrast, our approach is a more dynamic approach, which is performed during the real-time networking; in addition, no prior knowledge of the traffic profile is needed.

4 Will-be-set-by-IN-TECH

(i) Enhance the capability of the node to receive and transmit concurrently by ensure that distinct channel should be reserved for the reverse and forward routing entry for each

(ii) Integrated the channel assignment and distribution with the reactive gateway discovery process in order to efficiently utilize the limited number of non-overlapping channel and

(iii) Developed a hybrid interface assignment that reduces the packet collision for the gateway traffic due to the broadcast nature of the wireless medium and existing of local traffic. This done by proposed static and dynamic channel assignment. Static interface assign to static channel and used to support the local traffic while the dynamic interface only assign to active node during the gateway discover process. This interface used to

(iv) Developed channel assignment that simple (reduce the channel assignment complexity) and independent of any particular profile such as traffic, interference, and topology

The remainder of the chapter is organized as follows. Section two discusses relevant work. The AODV-MRCR protocol is explained in section three. In section four, we provide the details of our simulation environment. Simulation results and their analysis are presented

A major problem facing multi-hop wireless networks is the interference between adjacent links. The throughput of a single-radio single-channel wireless network has been studied in [37]. The authors formalized it as a multi-commodity flow problem with constraints from conflict graph, which is NP hard, and gave an upper bound and a lower bound of the problem. There have been many studies on how to assign limited channels to network interfaces in a multi-radio multi-channel wireless mesh network as to minimize interference and maximize throughput. They differ in several assumptions made in WMNs, and therefore in the models

One approach assumes a known traffic profile in the network, because the aggregate traffic load of each mesh router changes infrequently. The authors of [26] proposed an iterative approach to solve the joint routing and channel assignment problem. Heuristic techniques are used to estimate the traffic load in each link. The algorithm starts with an initial estimation of the expected traffic load and iterates over both channel assignment and routing until the bandwidth allocated to each virtual link matches its expected load. While this scheme presents a method for channel allocation that incorporates connectivity and traffic patterns, the assignment of channels on links may cause a ripple effect whereby already assigned links have to be revisited, thus, increasing the time complexity of the scheme. Moreover, this approach is performed during the network plan and assumes that the traffic profile is known. The centralized flow-based and rate channel assignment algorithm is proposed in a paper by [3]. The agreed heuristic algorithm is used for channel assignment rate. [6] enhances the [26] centralized algorithm to support automatic and fast failure recovery. The failure recovery mechanism is located at the gateway and all nodes send periodic messages to the gateway. In the case where the gateway does not receive a message during a period of time from node *x*, it deletes the corresponding information, node id, position, rate, and then runs

node involve in path establish process during the gateway discovery process.

establish high throughput paths for the gateway traffic.

supported gateway traffic.

in section five, with concluding remarks in section six.

profile.

**2. Related works**

and related solutions.

Other studies assume that the traffic profile of each mesh router is not known, and usually consider channel assignment and routing separately. The authors of [25] assumed that the traffic from the Internet gateway to clients is dominant, and thus proposed distributed channel assignment based on spanning tree topology, where the gateway is the root of the spanning tree. The protocol dedicates one interface channel for communication with its parent node on the tree, and the other interfaces are configured as children for communication with their child nodes. Hence, the protocol divides the node interfaces into two subsets - downlink and uplink interfaces. The uplink interfaces are used to connect the node with its parent node while the downlink is used to connect the node with its child nodes. The node can only switch its child. For channel assignment, the channel assignment strategy starts from the root of the tree. Each node switches its parent interfaces to the parent node child interface and selects a new channel for its child interfaces. One drawback of this protocol is that it only considers the common traffic where data are transmitted from the source to gateway and vice versa.

Multi-channel routing protocol (MCR) [17] the peer-to-peer traffic was assumed to be dominant in the network. The authors first constructed a k-connected backbone from the original network topology, and then assigned channels on the constructed topology. The MCR classified the node interfaces into fixed or switchable interfaces. The protocol assigns a fixed channel to the fixed interface for communication between neighbors, and the remaining interfaces are considered as switchable interfaces. When a node wants to communicate with others, it looks in its table to find the destination's fixed channel and switches one of the switchable interfaces to that channel. To exchange fixed channels between neighbors, MCR uses a "hello" message to carry the fixed channel information. However, this protocol may not work well in a multi-flow transmission because of high switching interfaces and because it does not utilize all the non-overlapping channels as the static channel assignment uses.

Although there are many distributed solutions proposed in literature [7, 9, 16, 23, 25, 38]. In [16], the authors proposed the Local Channel Assignment (LCA) algorithm, which adopts a tree-based routing protocol for common traffic similar to Hyacinth. The LCA algorithm solved the Hyacinth interface-channel assignment conflict problem which is caused when a parent switches to the least load channel that may be in use by one of its children. The interface-channel assignment problem may cause recursive channel switching and delays. LCA solved this problem by dividing the non-overlapping channel into groups and making each parent interface belong to one group different from its child interface group. The paper of [7] proposed a distributed joint channel assignment and routing protocol for multi-radio multi-channel ad hoc network. The scheme dedicates one interface for the control message and another interface for data transmission. The control interface is assigned to a common channel while the data interfaces could work as a fixed or switchable interface based on the receiving call direction. However, in this approach, the control interface becomes the bottleneck, especially in high-density networks.

In the paper of [9], the authors proposed a hybrid multi-channel multi-radio wireless mesh network architecture, which combines the advantages of both static and dynamic channel

#### 6 Will-be-set-by-IN-TECH 232 Wireless Mesh Networks – Effi cient Link Scheduling, Channel Assignment and Network Planning Strategies High Throughput Path Establishment for Common Traffic in Wireless Mesh Networks <sup>7</sup>

allocation strategies. The architecture is similar to Hyacinth architecture [25]; it classifies the interface to work as a fixed interface or a switchable interface. The protocol only considers one interface to work as a switchable interface. This interface has the ability to switch channels frequently, while the remaining interfaces are considered as fixed interfaces that work on fixed channels. The channel allocation of static interfaces aims at maximizing the network throughput from end-users to the gateway, while the dynamic interface is used to communicate with the neighbor node that has a different fixed channel on-demand fashion. Two dynamic interfaces that are within radio transmission range of each other are able to communicate by switching to the same channel when they have data to transmit.

AODV. The source node needs to ensure the channel selection by sending two messages. The first message is to inform neighbors about the selected channel and the second message is sent by the neighbors to confirm the channel. In case the channel is in use, the node should be waiting until the channel is free or selects a new channel. However, such an approach may not work well in WMNs where most of the traffic is directed toward the gateways and must

High Throughput Path Establishment for Common Traffi c in Wireless Mesh Networks 233

**3. Multi-radio ad hoc on-demand distance vector routing with channel**

is proposed to establish high throughput paths for the gateway traffic in WMNs.

the gateway path, one link for the forward path and other for reverse path.

range *m* − *k*/2 channels during the RREP\_I message.

as "unused channels". The *i* available interfaces at each node can be classified as:

The Multi-radio ad hoc on-demand distance vector routing with channel reservation scheme (AODV-MRCR) protocol is a multi-radio on demand distance vector routing protocol, which

Our scheme uses on-demand reactive routing protocol to distribute the reserved channels list among all the nodes along the path from the source to the gateway. The source node that does not has fresh route to the gateway, it sends RREQ with flag set to one which means only the gateway can reply this message. Once the gateway receives a new RREQ\_I, it selects a reserved channel list and attaches to the RREP\_I message. The message is sent back to the source node. During the RREP\_I stage, each intermediate node selects its recommend channel based in the hop count index[20]. The intermediate node reserves at least two interfaces for

We assume that *m* channels are available that can be used in a wireless area without interfering. In addition, *k* channels of available channel are statically assigned to *i* interfaces, half of *k* channels are used as "used channels", and the *m* − *k*/2 channels will be considered

• Fixed interfaces: Some *n* of the *i* interfaces at each node are assigned for long intervals of time to *k* channels, we designate these interfaces as "fixed interfaces", and the corresponding channels as "used channels". These interfaces are used to keep the network connectivity as well as support the local traffic. Therefore, they are not allowed to switch. • Switchable interface: The reaming *i* − *n* interfaces are switched to the selected channels for long intervals of time. These interfaces are assigned to a channel that is selected from the

For example, if the mesh router has four interfaces, two channels of twelve (1, 2) will be considered as used channels. The reaming channels will be considered as unused channels. Moreover, the interfaces one and two will be considered as fixed interfaces while the interfaces three and four are switch- able interfaces. The scheme consists of two parts. The first part is carried out at the gateway, which is used to reserve a unique list of channels for each RREQ\_I received at the gateway. The second part is carried out when the intermediate nodes along the path back to the source receive the RREP\_I message. Following is a clarification of the

Channel reservation scheme is carried out into two stages. First stage is carried out by the gateway, which is used to reserve a unique list of channel for each received RREQ\_I message,

pass through mesh routers.

**reservation scheme**

procedures.

**3.1. Channel reservation scheme**

In [23], the authors proposed a learning based approach for distributing channel assignments. It uses a learning based algorithm to determine the best channels to assign its own interfaces based on collecting information from the neighbor nodes. Hence, each mesh node periodically sends a "hello" message in order to discover its neighbors and the channel usage in its neighborhood. The algorithm achieves effective channel usage, and also adapts well to the change of network topology. [38] proposed a distributed channel assignment for uncoordinated WMNs to minimize the interference with adjacent access points. The algorithm assigns the least interference channel to the access point interference according to the gathered channel information from neighboring access points and associated clients. Both the protocols discussed earlier assign channels from node to node, and each node in the WMNs assigns a fixed channel, which makes it different from our approach. In our approach, channel assignment is based on data flow such that a channel is only assigned to a node if it has data to send or forward to the gateway.

The authors of [34] and [7] proposed algorithms to minimize network interference. The first one is interference-aware because it visits the links in decreasing order of the number of links falling in the interference range and it selects the least used channel in that range. Assuming the set of connection requests to be routed, both an optimal algorithm based on solving a Linear Programming (LP) and a simple heuristic are proposed to route such requests, given the link bandwidth availability as determined by the computed channel assignment. The algorithm considers minimum-interference channel assignments that preserve k-connectivity. The algorithm proposed in [7] uses a genetic approach to find the largest number that makes the whole network connected while minimizing network interference. However, such approaches only focus on minimizing the network interference that may decrease the network connectivity. In contrast to the above mentioned approaches, our approach is based on eliminating the interference for the common traffic on WMNs while maintaining network connectivity. Besides static channel assignment algorithms, which assign channels to interfaces without change for a long time, there have been several dynamic channel allocation algorithms proposed, which allow interfaces to switch channels frequently.

The authors of [29] proposed an on-demand channel allocation protocol in a wireless mesh network, where each node has two interfaces. In their framework, one interface of each node is devoted to controlling channel negotiation only while the other interface is used for data transmission. On the other hand, the frameworks proposed in [30] and [4] do not require a separate control interface, and the channel negotiation happens on the same interface for data transmission.

The Channel Assignment Ad hoc On-demand Distance Vector routing (CA-AODV) [12], has been proposed to assign channels within K hops in an ad hoc network, allowing for concurrent transmission on the neighboring links along the path and effectively reducing the intra-flow interference. Similar to CA-AODV, [33] proposed to join the channel assignment with the AODV. The source node needs to ensure the channel selection by sending two messages. The first message is to inform neighbors about the selected channel and the second message is sent by the neighbors to confirm the channel. In case the channel is in use, the node should be waiting until the channel is free or selects a new channel. However, such an approach may not work well in WMNs where most of the traffic is directed toward the gateways and must pass through mesh routers.
