In this paper, we consider the dynamic content caching issue in the cache-enabled Internet of Things (IoT) systems. For real-time applications in cache-enabled IoT systems, it is imperative to design dynamic content caching schemes to reduce the energy consumption of sensors and improve the freshness of information at users. We first design a dynamic content caching procedure for a cache-enabled IoT system with limited cache capacity and express the evolution of the Age of Information (AoI) at both the edge caching node and each user. Then, we formulate the dynamic content caching problem as a Markov Decision Process to minimize the expectation of a long-term accumulative cost, which jointly considers the average AoI of users and the energy consumption of sensors. To solve this problem, we propose an actor-critic based caching algorithm without prior knowledge of users’ content demands. The numerical results show that the proposed algorithm can achieve lower average AoI and energy consumption than other baselines.
Dynamic Content Caching Based on Actor-Critic Reinforcement Learning for IoT Systems
2022-09-01
832205 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Europäisches Patentamt | 2020
|Actor-Critic Reinforcement Learning for Control With Stability Guarantee
BASE | 2020
|Actor-Critic Traction Control Based on Reinforcement Learning with Open-Loop Training
BASE | 2022
|