| Popis: |
In this paper, we consider a wireless network consisting of a base station (BS) that is serving multiple real-time traffic streams forwarding information updates to their destinations in order to sustain the freshness of information for time-critical applications. Since the wireless channels may be unreliable due to the impurities of the propagation environments, such as deep fading, blockages, etc., we integrate a reconfigurable intelligent surface (RIS) to the wireless system in order to mitigate the propagation-induced impairments, enhance the quality of the wireless links, and ensure that the required freshness of information is achieved for these real-time applications. For this network set-up, we investigate the joint optimization of the traffic streams scheduling and the RIS phase-shift matrix with the goal of minimizing the long-term average Age of Information (AoI). The formulated optimization problem is a mixed-integer non-convex optimization problem, which is difficult to solve. To circumvent the high-coupled optimization variables, and with the aid of bilevel optimization, we decompose the original problem into an outer traffic stream scheduling problem and an inner RIS phase-shift matrix problem. For the outer problem, owing to its complexity and stochastic nature of packet arrivals, we resort to deep reinforcement learning (DRL) solution where the traffic stream scheduling is modeled as a Markov Decision Process (MDP), and Proximal Policy Optimization (PPO) is invoked to solve it. Whereas, the inner problem that determines the RIS configuration is solved through semi-definite relaxation (SDR). Finally, we show through extensive simulations that our approach evaluates the combined impact of scheduling policy and RIS configuration on the long-term average AoI, where we demonstrate its superiority against other baseline schemes. |
