Accessibility statement

Application of Reinforcement Learning on Medium Access Control for Wireless Sensor Networks

Wireless Sensor Network (WSN) is a self-organised network which consists of a large number of randomly deployed nodes. It has a wide range of applications in industry, security, military, medical treatments and environmental research. Medium Access Control (MAC) protocols deal with how to schedule transmission and reception while sharing the same channel resource, and they have significant impact on energy efficiency and network life-time.

According to the energy constraints of WSNs, MAC protocols need to minimise energy cost associated with idle listening, overhearing and control packets while maintaining necessary throughput and delay performance. This project aims to develop energy-efficient MAC protocols with low complexity based on ALOHA and Reinforcement Learning.  Reinforcement Learning offers the potential for nodes to learn the most appropriate times to transmit on a shared channel, avoiding contention and resulting in almost perfect slot selection, thereby improving the energy efficiency.

The protocols are being simulated in OPNET alongside other protocols for comparison.

Key objectives

  • To understand the special requirements of MAC protocols for WSNs.
  • To understand the advantages and disadvantages of ALOHA and other existing MAC protocols.
  • To apply Reinforcement Learning to Slotted ALOHA and compare with other protocols.

Members

  • Yi Chu
  • Paul Mitchell
  • David Grace
  • Tim Clarke

Dates

  • October 2009 to
    September 2013

Research