Development of an AI Agent-Based Hierarchical Swarm Autonomous Control System (H-SAC) for Resilient Heterogeneous Mobility in Precision Agriculture

Abstract

Precision agriculture increasingly depends on autonomous aerial platforms for pollination, crop scouting, and site-specific spraying. In practice, however, field deployment remains constrained by operational stochasticity, including battery depletion, wind-induced mission interruption, canopy occlusion, and the limited ability of a single platform to sustain continuous work. To address these limitations, this paper presents a Hierarchical Swarm Autonomous Control (H-SAC) architecture that coordinates heterogeneous aerial and ground agents through a three-layer AI-agent framework composed of Observer, Edge AI, and Worker layers. The proposed system integrates edge-based perception, priority-aware task allocation, and a heterogeneous mobility handover mechanism that transfers mission execution from drones to ground robots when environmental conditions degrade. The framework was implemented in a ROS2-Gazebo simulation environment representing a dense pear orchard and evaluated under nominal and disturbed operating conditions. Results show that H-SAC reduces total operation time by 35.4% relative to a single-drone baseline, achieves 98.2% pollination coverage, and maintains task continuity during an 8\m/s wind disturbance by triggering sub-second aerial retreat and coordinated ground takeover. These findings indicate that hierarchical multi-agent orchestration can substantially improve robustness, continuity, and field-level efficiency in precision agriculture.