A Case Study for Vulnerable Road Users Protection: Systematic Approach for Managing and Scheduling C-ITS Workloads in Heterogeneous Clusters

Safety-critical C-ITS applications-such as Vulnerable Road User (VRU) protection-require low latency, as mandated by international standards. Current communication infrastructures often fail to meet these stringent requirements due to end-to-end cumulative delays in data collection, cloud-based decision-making, and response delivery. While edge computing reduces latency, achieving predictable performance across the edge-cloud continuum requires intelligent workload orchestration that dynamically adapts to system conditions. This paper introduces a network-aware orchestration system that replaces the native scheduler of a container orchestration platform to enable latency- and energy-sensitive workload placement in heterogeneous clusters, considering processing and communication metrics. A Decision Assistance Service (DA Service) is integrated, which leverages empirical delay and energy profiles along with real-time network metrics exported from cluster nodes. The scheduling problem is modeled as a binary integer linear program that dynamically balances latency-sensitive and energy-efficient objectives through tunable trade-offs. The proposed approaches are validated through real-world deployments at multiple intersections equipped with IP cameras and DSRC-based roadside units. Real-world experimental results across three scenarios show that our system consistently meets strict latency constraints while optimizing energy consumption. These findings highlight the critical importance of integrating real-time network and device metrics into orchestration logic to ensure responsiveness and predictability in distributed C-ITS environments.