Visual Insights into Agentic Optimization of Pervasive Stream Processing Services
Researchers demonstrate AI agents that autonomously optimize competing streaming services on resource-constrained edge devices.
Researchers from TU Wien (Boris Sedlak, Víctor Casamayor Pujol, Schahram Dustdar) present a visual platform for agentic optimization of pervasive stream processing. It tackles fluctuating demand and limited edge resources by deploying scaling agents that explore each service's unique action space. The agents learn optimal policies to dynamically adjust service execution, preventing co-located services from cannibalizing each other's resources while maintaining low latency for applications like smart cities.
Why It Matters
Enables reliable, low-latency IoT and smart city applications by intelligently managing scarce compute at the network edge.