Channel Allocation in Wireless Sensor Networks

Wireless sensor networks (WSN) are a recent class of networks able to accurately monitor our daily environment. The network is composed of tiny devices known as wireless sensors or motes. Each node has a microprocessor, a memory, a (multichannel) radio, a battery, and one or more sensors such as temperature, humidity, or sound sensors. The figure below is a schematic representation of a wireless sensor network where sensor nodes form a multi-hop routing tree to the base station.

Many communication protocols allow nodes to use different frequencies (i.e., channels) for transmission, in order to improve network throughput. However, the sender and receiver need to select the same channel for proper communication. Moreover, nearby nodes communicating on adjacent channels can cause radio interference. It is, therefore, a challenging task to make nodes learn to coordinate their transmissions in a distributed manner.

In addition, many multichannel protocols implicitly assume that the packet reception is similar across all the channels, which is not true in reality. Machine learning techniques would allow to extract sufficient information about the quality of each channel in order to select the best frequency for communication.

The goal of this thesis is to design machine learning techniques for dynamic channel allocation in order to improve message throughput in wireless sensor networks. The study will be done by software simulation and will be validated by hands-on experiments on real testbeds.

For questions concerning this thesis proposal, do not hesitate to contact both supervisors.

Supervisors: Kieu Ha Phung and Mihail Mihaylov (Mike)

Promoters: Kris Steenhaut and Ann Nowé