A Multi-Agent Reinforcement Learning-Based Optimized Routing for QoS in IoT

The Routing Protocol for Low power and lossy networks (RPL) is used as a routing protocol in IoT applications. In an endeavor to bring out an optimized approach for providing Quality of Service (QoS) routing for heavy volume IoT data transmissions this paper proposes a machine learning-based routing algorithm with a multi-agent environment. The overall routing process is divided into two phases: route discovery phase and route maintenance phase. The route discovery or path finding phase is performed using rank calculation and Q-routing. Q-routing is performed with Q-Learning reinforcement machine learning approach, for selecting the next hop node. The proposed routing protocol first creates a Destination Oriented Directed Acyclic Graph (DODAG) using Q-Learning. The second phase is route maintenance. In this paper, we also propose an approach for route maintenance that considerably reduces control overheads as shown by the simulation and has shown less delay in routing convergence.