Towards Agricultural Vehicle Autonomy: Adaptive PID Control for Precise Path Tracking with Visual-Inertial Localization

Agricultural production relies on manual machinery, resulting in high labor dependency and low efficiency. Recent advancements introduce autonomous tractors, yet challenges persist in path-following control due to complex dynamics and reliance on high-cost GPS systems. This paper presents a novel approach for developing autonomous tractors utilizing a cost-effective visual-inertial localization system. The proposed system integrates multi-view cameras and IMUs for reliable pose estimation and employs an adaptive PID controller based on a first-order plus time-delay model to ensure high-precision path tracking. The modular design allows minimal retrofitting of existing machinery. Experimental results demonstrate the system's feasibility for high-precision, cost-effective agricultural automation under real-world conditions.