Adaptive Scheduling for Real-Time Safety in Smart Intersections Using the DECICE Framework

The Smart Intersection with VRU Detection use case, led by Marmara University, showcases low-latency, real-time safety improvements in connected transportation through multi-intersection coordination and adaptive traffic management. Leveraging the DECICE framework, it ensures efficient computational orchestration, dynamic adaptation, and smart scheduling for edge-cloud applications. Camera frames are processed on dynamically scheduled nodes to meet strict latency constraints, ensuring safety responses remain timely. Marmara’s infrastructure supports dynamic workload deployment and latency optimization, with scenarios demonstrating adaptive scheduling. In Scenario 1, increased workload metrics were observed on occupied nodes, conserving energy at the cost of slightly higher inference times. In Scenario 2, artificially increased latencies led to idle nodes being activated, maintaining tolerable delays. This illustrates how the DECICE framework’s scheduler adapts to changing conditions for optimal performance.

Author(s): Yavuz Selim Bostanci, Marmara University