A SURVEY ON CHANNEL ASSIGNMENT FOR MULTI-RADIO MESHED NETWORKS

Maria Dolores Salmerón Pérez

INTRODUCTION: WIRELESS MESH NETWORKS

Wireless Mesh Networks (WMNs): mix between two wireless networks topologies: ad-hoc networks and structural topology.

Differences between ad-hoc networks and WMNs Topology changes Mobility Scenario of application

Important research area Increasing the coverage area of the network Larger number of access points

INTRODUCTION: WIRELESS MESH NETWORKS

WMNs and Ad hoc networks: Normally single radio systems

Disadvantages: Lower throughput Nodes with half-duplex mode Many channel changes due to dynamic network traffic Interference from external sources Shared Spectrum

Solution: Multiradio systems (most of the current WMN deployments adopt the multi-radio multi-channel architecture)

INTRODUCTION: WIRELESS MESH NETWORKS

Based on the standards: IEEE 802.11 a (5 GHz) / IEEE 802.11 b (2.4 GHz)

IEEE 802.11a/b standards provide respectively 12 and 3 non-overlapped frequency channel

Ability to use multiple channels substantially increase the effective bandwidth available to wireless networks

INTRODUCTION: WIRELESS MESH NETWORKS

In the standard IEEE 802.11b there are 12 available channels but only the channels 1, 6 and 11 have low interference and/or overlapping with each other.

PROBLEM FORMULATION

A pair of nodes that use the same channel and within interference range may interfere with each other’s communication

Figure (b) is an example of channel assignment for figure (a) that minimizes the number of interfering links for nodes A and B

SOLUTION: CHANNEL ASSIGNMENT

For any mesh network, there are multiple ways to assign channels to the radios

Solution: Find a proper channel assignment in every node of the network

CA algorithms

Goals: Increased bandwidth Performance improvement

SOLUTION: CHANNEL ASSIGNMENT

Classification of Channel Assignment approaches for MR-MC WMNs :

Centralized: central control Distributed: no central control

SOLUTION: CHANNEL ASSIGNMENT

Centralized algorithm: CLICA (Connected Low Interference Channel Assignment)

Polinomial time greedy heuristic Traffic independent Based on the connectivity graph and on the

conflict graph Every node is associated with a priority Coloring decisions are made node by node basis

in the order of this priority Reducing interference

SOLUTION: CHANNEL ASSIGNMENT

Example of CLICA algorithm:

SOLUTION: CHANNEL ASSIGNMENT

Distributed algorithm: DCA (Distributed Coloring Algorithm)

Clustering algorithm is applied previously to make set of nodes.

Distributed Clustering Algorithm that find the cluster-head

Cluster-head is the responsible of the graph coloring i.e. of the channel assignment

Every cluster-head has a priority or weight and knows the priorities of its neighbours nodes

Cluster-head colors its cluster only if highest priority cluster-head neighbors have colored completely their clusters

SOLUTION: CHANNEL ASSIGNMENT Example of DCA algorithm:

There are 3 types of messages that determine the action of every node:

CA: allows any node to know its colorUP: updates the color of one nodeINF: inform about forbidden colors

SOLUTION: CHANNEL ASSIGNMENT

Distributed algorithm: MIX (Minimum Interference Channel Selection)

Clustering algorithm is applied previosly to make set of nodes.

Highest Connectivity Cluster algorithm that find a cluster-head

MIX is based on minimizing co-channel interference between clusters

Access points have two interfaces: First interface: for inter-cluster communication Second interface: for intra-cluster communications

using the channel selected by the cluster-head through MIX algorithm

SOLUTION: CHANNEL ASSIGNMENT Example of MIX algorithm:

CONCLUSIONS

DCA is an algorithm designed for devices on a single radio interface and a relatively low consumption

MIX algorithm is the most appropriate algorithm for a larger size of network and the throughput is better than DCA

CLICA has some limitations but compared with the distributed algorithms the performance is a bit better

CONCLUSIONS

Centralized algorithms usually have in the most of the cases a better performance because they used entire network information

Centralized algorithms: nearer throughput to the optimal

Problem of the algorithms NP-hard problems (difficult to find an optimal solution and exponential runtime )

Heuristics approximation modes (not optimal results)

QUESTIONS