PROJECT TITLE :

Multi-hop Deflection Routing Algorithm Based on Reinforcement Learning for Energy-Harvesting Nanonetworks

ABSTRACT:

Nanonetworks are made up of nano-nodes that interact with one another, and the size of these nano-nodes can range anywhere from several hundred nanometers to several micrometers. The design of routing protocols for nanonetworks is extremely difficult due to the extremely limited computational resources of nano-nodes, the fluctuations in their energy caused by energy harvesting processes, and their very limited transmission range at Terahertz (THz)-band frequencies (0.1-10 THz). In this paper, a multi-hop deflection routing algorithm that is based on reinforcement learning is proposed. The goal of this algorithm is to dynamically and efficiently explore the routing paths that are taken while packets are being transmitted. First, new routing and deflection tables are implemented in nano-nodes. These tables allow nano-nodes to deflect packets to other neighbors in the event that route entries in the routing table contain invalid information. Second, in order to update the tables, one forward updating scheme and two feedback updating schemes that are based on reinforcement learning have been designed. These schemes are referred to as on-policy and off-policy updating schemes. In the end, exhaustive simulations are carried out in networks simulator-3 in order to analyze the performance of MDR-RL utilizing various updating policies and to compare this performance with that of other machine learning routing algorithms based on Neural Networks and Decision Tree. According to the findings, the MDR-RL has the potential to raise the percentage of successfully delivered packets, as well as the total number of successfully delivered packets, while simultaneously lowering the average number of hops per packet.