Energy Efficient Bandwidth Allocation and Routing in Electromagnetic Nano-Networks via Reinforcement Learning

Electromagnetic nano-networks operating in the THz band offer a promising solution for enabling communication among many nanoscale devices. However, the inherent limitations of nano-nodes, such as restricted energy and processing resources and short communication range, pose significant challenges for efficient data transmission. While prior research has explored Reinforcement Learning (RL) for optimising traffic routing in electromagnetic nano-networks, this paper proposes a novel approach that jointly optimises routing and sub-channel bandwidth allocation to minimise network energy consumption using RL. We leverage the Q-learning algorithm to develop a dynamic single-hop or multi-hop routing scheme that considers each node's location, energy storage capability, and the available sub-channel bandwidth. Our model formulates a reward function that balances these multiple objectives and enables the selection of optimal transmission policies for each nano-node. Our findings suggest carefully choosing the number of hops and increasing bandwidth in sub-channels can lead to substantial energy savings in nano-networks.