Experimental Evaluation of a Remote Monitoring-Based Guidance System for Semi-Autonomous Tractors in Paddy Fields

Highlights The proposed guidance system enables road navigation and field operations across multiple paddy fields. Multi-camera obstacle detection approach detects obstacles in predefined hazard zones surrounding the tractor. Remote supervision enables overseeing of tractor operations and surrounding safety. Abstract. The increase in global food demand coupled with persistent labor shortages has spurred the development of autonomous agricultural equipment such as autonomous tractors. However, implementing fully autonomous tractors in complex agricultural settings presents unique safety, reliability, and adaptability challenges. To address these issues, a remote monitoring-based guidance system (RMGS) for semi-autonomous tractors is proposed. The RMGS integrates path-planning algorithms for on-road driving and in-field tasks, a multi-camera obstacle detection method using YOLOv8s and inverse perspective mapping (IPM), and remote supervision that enables human oversight. The system was evaluated via field experiments in four paddy fields that collectively covered a total area of 1500 m 2 . The results demonstrated path tracking accuracy, with root mean square errors (RMSEs) of 0.05 and 0.12 m for straight and curved road segments, respectively. In-field performance yielded RMSEs of 0.04 m during agricultural tasks and 0.22 m for turning maneuvers. The obstacle detection system activated the emergency stop mechanism with 100 % accuracy, an average stopping distance of 3.6 m, and a response time of 0.4 s. The remote supervision facilitated reliable monitoring, delivering emergency notifications within an average of 0.9 s. Despite the demonstrated effectiveness, this study has limitations. The experiments were conducted on clear days during daylight hours. Thus, further validation is required under varied environmental conditions. Moreover, the RMGS was tested with a single supervisor managing one tractor, necessitating further research on scalability for multi-tractor operations. This study demonstrates the effectiveness of RMGS for semi-autonomous tractor operation with human oversight. Keywords: Obstacle detection, Paddy fields, Remote supervision, Semi-autonomous tractors.