CO-QLR: Cooperative Q-Learning based Routing for IoT Networks

Low-power and Lossy Networks (LLNs) are characterized by constrained nodes with limited power, processing capabilities, and memory, leading to challenges such as high packet loss, low transmission rates, and network instability. While the standardized IPv6 Routing Protocol for LLNs (RPL) addresses some of these issues, it falls short in dynamic IoT environments, particularly under congestion. To enhance routing efficiency in LLNs, this paper proposes CO-QLR (Cooperative Q-Learning-based Routing), a novel distributed routing protocol that leverages cooperative reinforcement learning. Unlike independent Q-learning-based approaches, CO-QLR enables each node to share learned metrics from its Q-table with neighboring nodes, thereby improving load balancing and reducing packet loss. Simulation experiments are conducted to evaluate CO-QLR’s performance against conventional independent Q-learning-based routing, demonstrating its effectiveness in reducing message loss rate. This study contributes a cooperative learning framework for routing in constrained networks, showing potential for enhancing the resilience and efficiency of LLNs in diverse network topologies.