Optimizing Latency and Energy Efficiency in Edge Computing with Reinforcement Learning and TinyML
S. Saraswathi
Abstract
Now, a new solution has stepped forward which is promising to offer just that in low latency real time smart networks especially in areas such as Wireless Sensor Networks. This paper puts forward an AI-Driven Adaptive Edge Computing Framework (AEAO) to minimize data processing delay based on real-time resource provisioning and smart task allocation. The framework also initiates lightweight AI models like the TinyML to do inference at certain regions; it also employs RL to manage the resources flexibly. Also, Federated Learning (FL) increases the privacy of the system since models are trained across distributed edge nodes without aggregating sensitive data. The experimental setup of our work assesses the efficacy of the proposed framework in the model healthcare and industrial IoT settings based on the corresponding parameters, including latency, throughput, and energy consumption. Analysis with other methods suggests that delays in processing have decreased, and the accuracy of decisions is higher compared to conventional cloud-oriented and immovable edge computing models. This research captures the possibility of synchronizing computing at the network edge with AI in tackling contemporary challenges of real-time, secure, and efficient data consumption in intelligent networks