DRL-Based Adaptive Edge Caching for Enhanced Content Delivery
Dr. Akula Suneetha, K. Ratna Babu, Inakoti Ramesh Raja, V. Nagagopiraju, Eppili Jaya, K. B. Glory, Naga Lakshmi · International Journal of Basic and Applied Sciences · 2025
The number of devices using the internet is growing fast. More people are using smartphones, smartwatches, and other smart devices. These devices send and receive large amounts of data. This is causing heavy traffic on mobile networks. Because of this, users sometimes experience slow responses and de-lays in getting content. To fix this problem, one idea is to use edge caching. When users request data, it is delivered faster from nearby servers instead of faraway cloud centers. This helps improve user experience and reduce network load. This lead to low hit rates, which means users still need to fetch data from far servers. We proposed a smart caching system. It is called ICE (Intelligent Caching at the Edge). ICE uses a technique called Deep Reinforcement Learning (DRL). DRL is a form of machine learning. It helps the system learn from past experiences and make better decisions over time. ICE decides which content should be stored at the edge, and it is removed. ICE uses a special popularity model. It is based on Newton's Law of Cooling. This model predicts how popular a piece of content will be over time. The system uses a type of decision-making model called a Markov Decision Process (MDP). This helps ICE choose the best action in each situation—whether to cache, replace, or skip content. The system is designed to improve the cache hit rate and reduce the time and energy needed to fetch data. The paper proposes a second system called DCCC (Distributed Cooperative Caching with Coordination). DCCC helps multiple edge nodes work together. It uses a two-layer caching setup and spreads content requests across nodes. This improves caching efficiency and uses storage space better. We test ICE and DCCC using many experiments. DCCC improves caching by using resources from many nodes. ICE learns to store the right content using DRL. DCCC makes caching more efficient by coordinating across many edge nodes.