English Abstract
Abstract :
The growing presence of Internet of Things (IoT) devices requires efficient and
reliable data transmission protocols. However, traditional routing protocols struggle in dynamic and resource constrained IoT networks due to factors like decentralization,
limited energy, and frequent topology changes. This thesis proposes a novel routing
protocol, modified Energy-Efficient Routing based on Reinforcement Learning
(mEER-RL), specifically designed to address these challenges and enhance
performance in both fixed and mobile IoT deployments. mEER-RL is built over the
foundation of EER-RL, a cluster-based, RL-driven protocol focused on energy
efficiency. mEER-RL contains a significant modification to address node mobility, a
crucial factor often neglected in existing cluster-RL based protocols for IoT. By
integrating mobility considerations into the RL agent's reward function, mEER-RL
dynamically adapts its routing decisions to network conditions, improving Packet
Delivery Ratio (PDR) and overall network reliability. This thesis presents a survey of
most types of cluster-based routing protocols built for IoT networks. RL techniques are
also reviewed to be applied for the routing protocol. The thesis then comprehensively
evaluates mEER-RL's performance through simulations, comparing it against the
established base protocol LEACH and EER-RL. The evaluation focuses on crucial
metrics like network lifetime, energy consumption, packet delivery ratio, and the
impact of node mobility. The results demonstrate that mEER-RL effectively balances
energy efficiency, reliability, and adaptability, making it a promising solution for
routing in complex and dynamic IoT networks. The key contributions of this work
include enhanced reliability, increased PDR, enhanced adaptability, and improved
energy efficiency for both fixed and mobile IoT environments by optimizing routing
paths and minimizing data transmission through improved multi-hop communication
techniques. For an advanced performance, compared to LEACH and EER-RL, mEERRL exhibits superior performance in terms of network lifetime, energy consumption,
and PDR.
