Energy Aware Controller Load Balancing Based on Multi-Agent Deep Reinforcement Learning for Software-Defined Internet of Things

Fluctuations in traffic within the Internet of Things (IoT) can affect the performance of the control plane. It is important to maintain stable control plane performance by load balancing strategies. To address the issue of controller load balancing in software‐defined Internet of Things (SD‐IoT), and meet the energy consumption requirements of nodes in the IoT during the adjustment process, a load balancing algorithm based on multi‐agent deep reinforcement learning (MADRL) is proposed. This approach models two critical factors: load difference and migration cost, and constructs a load balancing optimization problem based on these two factors. Subsequently, considering the dynamic changes in the state of the SD‐IoT, the load balancing problem is formulated as a Markov game process, and an algorithm is designed based on MADRL to solve this problem. Finally, the algorithm is validated based on real‐world topology, and a comparison is conducted from multiple perspectives including delay, load difference, energy consumption, and migration cost, demonstrating the effectiveness and advantages of the proposed algorithm.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.