ETD

• Edge-computing
• Reinforcement Learning
• Scalability
• Telemedicine

Patients with chronic diseases alternate between stable conditions and high-risk episodes that require continuous and accurate monitoring. Telemedicine systems rely on multiple battery-powered personal devices to collect and transmit health data over wireless networks, allowing clinicians to access updated patient information, especially in critical situations. However, frequent data collection and transmission can shorten device battery life, while an increasing number of monitored patients may cause network congestion and higher communication latency. To address these challenges, this thesis proposes an edge-computing solution that dynamically adjusts data collection and transmission frequency according to each patient's condition. A reinforcement learning algorithm based on a discrete Soft Actor-Critic policy is used to prioritize patients: critical patients are monitored and their data transmitted immediately to ensure low latency, while stable patients are monitored less frequently, reducing network traffic and extending battery lifetime.