Autonomous Sprinkler System with MAPE-K

Sprinkler is a key component for farms and gardens. However, conventional sprinkler systems have limitations: underwatering and overwatering under evolving weather, not considering plant conditions, and burden of manual scheduling. We developed Autonomous Sprinkler System to remedy the limitations. We acquire environmental contexts with loT sensors, cameras, and actuators, and apply MAPE-K for autonomic control. We implemented the Knowledge with a SVM classifier for inferring the situations and a Fast R-CNN model for determining plant condition. Our experiments show the system yields a considerable savings of water consumption and promotes the plant health, all without human users' interventions.