Enhancing real-time performance in mobile edge computing through age-aware deep reinforcement learning

Mobile Edge Computing (MEC) is growing fast. Many real-time applications use it to process information quickly. The freshness of information is very important. Age of Information (AoI) measures how fresh the data is. Lower AoI means the data is new. Higher AoI means the data is old. Many systems assume data is updated regularly. However, some MEC applications update data only when an event happens. Also, this data needs extra processing before it is useful. This paper redefines AoI to include processing time. The paper studies how to reduce AoI in MEC systems. The goal is to minimize AoI while managing system constraints. These constraints include bandwidth, energy and system resources. The problem is modelled as a Markov Decision Process (MDP). MDPs help in making decisions when future states depend on present actions. Reinforcement Learning (RL) can solve MDPs. However, traditional RL methods take a long time to learn. To speed up learning, the paper introduces Post-Decision States (PDSs). PDSs help the system use some known information. This makes learning faster and more efficient. Deep Reinforcement Learning (DRL) is used along with PDSs. DRL helps handle large and complex systems. The new method improves scalability and performance. The proposed method was tested in different scenarios. The results show that it works better than other methods. The new method reduces AoI more efficiently. It also uses less energy and resources. This makes it better for real-time applications. This paper improves AoI management in MEC systems. It uses DRL with PDSs to find better scheduling policies. The results show better performance compared to existing methods. This work can help future MEC applications process data faster and more efficiently.