Reinforcement Learning-Based Strategies For Achieving Long-Term Sustainability In IoT

In today's landscape, the demand for Internet of Things (IoT) devices have increased, leading to a vast network of heterogeneous devices communicating with each other. However, this connectivity poses significant challenges such as privacy concerns, accuracy issues, and energy inefficiency. Addressing these challenges requires innovative solutions, and one emerging technology poised to tackle them is the federated reinforcement learning framework. This framework offers several advantages for IoT devices, including increased learning speed and enhanced security through centralized model updates instead of raw data transmission. By incorporating federated learning with dueling DQN, our proposed technique offers improved accuracy by mit-igating the overestimation problem prevalent in other methods. In comparison to existing approaches, our Federated Dueling deep Q network technique demonstrates superior optimization in energy efficiency and delay while delivering enhanced accuracy. This combination of benefits positions it as a leading solution for the evolving needs of IoT device management and optimization.