Digital Twin Temporal Resource Allocation and Control Framework for IoT Networks Using TD3-LSTM

A Digital Twin (DT) system serves as a simulation tool designed to enhance the performance of real-time systems that are synchronized with digital twins, such as smart cities, autonomous vehicles, smart industries, and other applicable adaptive scenarios. This paper introduces the Digital Twin Temporal Resource Allocation and Control (DT-TRAC), a hybrid framework grounded in deep learning and reinforcement learning, aimed at enhancing performance and optimizing IoT networks, which are formulated as a Markov decision process (MDP). The proposed framework integrates the Twin Delayed Deep Deterministic Policy Gradient (TD3) with a Long Short-Term Memory (LSTM) network to accurately predict and adapt to the dynamics of networks in a partially observable environment, thereby reducing latency with improved performance. The proposed framework demonstrated improved performance metrics compared to baseline approaches with an accuracy of 95.5% and a latency of 19.5%. However, the proposed model exhibits an accuracy enhancement of 27%, from baseline to proposed model representing a significant advancement in dynamic environments using classification method on continuous control variables.