A latency-aware cloud–edge architecture for real-time injury risk early warning using wearable and training-load data

Injury risk monitoring in training and physical education settings increasingly relies on continuous streams of wearable signals and session-level workload records. These data sources are characterized by high sampling rates, strong temporal dependency, and strict latency requirements, which make conventional cloud-centric solutions insufficient for timely early warning. In view of this challenge, this paper presents a latency-aware cloud–edge architecture designed to support real-time injury risk early warning based on wearable and training-load data. The proposed architecture distributes data processing across device, edge, and cloud layers, enabling low-latency signal preprocessing and preliminary risk inference at the edge, while reserving the cloud for large-scale model training, historical analysis, and system-wide coordination. Experimental evaluations are conducted using simulated workloads that emulate realistic training scenarios. Experiment results demonstrate that the proposed cloud–edge architecture significantly reduces end-to-end latency and improves system scalability compared with cloud-only and edge-only baselines, while maintaining stable early-warning performance under varying data rates and network conditions.