سينا اسمعيلي

With the rapid growth of the Internet of Things (IoT), the importance of data freshness has become increasingly evident. Wireless communication networks based on Non-Orthogonal Multiple Access (NOMA) combined with Reconfigurable Intelligent Surfaces (RIS) can simultaneously enhance data freshness and overall system performance. The use of RIS in these networks not only aids in improving data transmission but also enables optimal multi-user transmission control through deep reinforcement learning algorithms. Together, these technologies play a crucial role in enhancing the efficiency of IoT systems, directly leading to improved performance. In this thesis, a NOMA network assisted by a reconfigurable intelligent surface is examined for collecting IoT device packets. Specifically, a novel model for eva‎luating packet freshness is presented. To minimize the average peak Age of Information (AoI), an optimization problem is formulated to jointly optimize the RIS phase shift matrix and packet service times. The Twin Delayed Deep Deterministic Policy Gradient (TD3) method is then employed to solve the non-convex problem, capable of handling a large number of continuous and high-dimensional variables. Extensive simulation results demonstrate that our proposed scheme outperforms conventional schemes. Simulations show that using the TD3 method and jointly optimizing the RIS phase shift matrix and packet service times can significantly reduce the average AoI. Comparisons of results with and without RIS usage indicate that the use of reconfigurable intelligent surfaces greatly facilitates the reduction of average peak AoI, leading to significant improvements in maintaining data freshness and faster responsiveness to environmental changes.

شبكه‌هاي ارتباطي بي‌سيم , دسترسي چندگانه غيرمتعامد , يادگيري تقويتي عميق , ديرينگي اطلاعات

Wireless Communication Networks , Non-Orthogonal Multiple Access (NOMA) , Deep Reinforcement Learning (DRL) , Age of Information (AoI)

Sina Esmaeili

Vesal Hakami