A Reinforcement Learning-Based Duty Cycle Adjustment Technique in Wireless Multimedia Sensor Networks

Abstract

Multimedia delivery support has recently been added to Wireless Sensor Networks (WSN) and has led to increased interest in Wireless Multimedia Sensor Networks (WMSN). WMSNs are expected to be crucial to the success of applications related to the Internet of Things (IoT), such as smart health, smart surveillance, smart homes, etc. Alongside their improved multimedia capabilities, WMSNs inherit WSN limitations such as energy and processing constraints. Additionally, WMSNs have significant Quality of Service (QoS) requirements, since multimedia delivery requires increased network performance in terms of bandwidth, latency, etc. Balancing energy efficiency and QoS is a fundamental challenge for WMSN users and operators alike. This paper proposes Reinforcement Learning based Duty Cycle (rlDC), an innovative learning-based scheme to adjust the duty cycle and contention window of WMSN nodes in order to meet energy efficiency and QoS targets. By employing rlDC, WMSN sensor nodes intelligently adapt their operation according to network delivery performance and application requirements. The proposed rlDC scheme was evaluated under different use cases in a simulation environment, and testing results show it outperforms other state-of-the-art duty-cycle-based protocols for WMSNs.

Keywords

• Wireless multimedia sensor networks
• energy efficiency
• Quality of Service
• duty cycle
• reinforcement learning
• medium access control