A Knowledge‐Augmented Wearable System for Interpretable Plant Health Monitoring

ABSTRACT In precision agriculture, modern sensing technologies can autonomously collect and process vast plant data using statistical or machine‐learning models. However, an “interpretability gap” persists: these outputs often remain black‐box correlations rather than interpretable physiological insights, which means users cannot validate their credibility, trace causal links to plant physiology, or translate them into actionable decisions. To bridge this gap, we propose a knowledge‐augmented wearable system for interpretable plant health monitoring (K‐Wear). By coupling a soft wearable sensing platform with a knowledge‐guided reasoning framework, K‐Wear leverages a large language model to transform raw sensing signals into mechanistic diagnoses and recommendations for proactive management. In simulated evaluations, the system achieved 99% diagnostic accuracy and first correctly detected plant stress on day 4.95 on average. In a real 14‐day potassium‐deficiency monitoring experiment, K‐Wear successfully identified early stress signals on day 4 and generated targeted intervention recommendations, enabling timely treatment and recovery of plant health.