Speaker: Abednego Wamuhindo Kambale
24 Ottobre 2025 | 14:30
DEIB - Aula Beta (Ed. 24)
On line by Teams
Contatti: Prof. Giovanni Quattrocchi
Sommario
On October 24, 2025 at 2.30 pm will we have a new seminar of DEEPSE Forum series titled "Reinforcement Learning-based methods for runtime management of AI applications in Smart Eyewears".The increasing adoption of Smart Eye-Wear (SEW) devices for AI-powered applications—such as object detection, augmented reality, and scene understanding—has introduced significant challenges in managing computational resources across resource-constrained wearables, edge servers, and cloud infrastructure. Dynamic variabilities in network throughput, cloud workload, and device mobility create unpredictable conditions that negatively impact Quality of Service (QoS) and Quality of Experience (QoE). This talk presents reinforcement learning approaches designed to address these challenges through adaptive task offloading strategies. The talk discusses how RL methods—progressing from tabular algorithms to hierarchical multi-agent systems—learn to balance energy consumption, execution performance, communication costs, and application accuracy despite uncertainties in the environment.