Digital Twin-Based Localisation Framework for Autonomous Agricultural Robots: A Multi-Sensor Fusion and Bayesian Estimation Approach

Résumé fourni par la source

Autonomous rice transplanting helps alleviate workforce shortages and boost agricultural output. Wheel slippage, varied topography, changing environmental conditions, and sensor ambiguity make flooded paddy field localisation difficult. DT-Loc is a Digital Twin-Based Localisation Framework for Tamil Nadu-based Japanese autonomous rice transplanting systems. Synchronised Digital Twin modelling, RGB-D imaging, LiDAR sensing, GPS, IMU, wheel odometry, Bayesian multi-sensor fusion, recursive state estimation, adaptive environmental mapping, and trajectory planning are integrated into the cyber--physical architecture. Experimental review compares EKF-SLAM, FastSLAM 2.0, ORB-SLAM3, LOAM, LIO-SAM, and Digital Twin-EKF using the publicly available SugarBeets2016 Agricultural Robot Dataset. DT-Loc reaches 99.08% localisation accuracy, reducing position RMSE to 0.112 m, mean absolute error to 0.087 m, and heading error to 0.91°. Digital Twin synchronisation achieves 98.73% environmental consistency with 22.61 ms update latency, while autonomous navigation achieves 99.26% row-following accuracy and 99.61% path completion. These results show that the proposed framework improves localisation reliability, environmental synchronisation, and navigation stability, making it an intelligent solution for precision rice cultivation workload reduction.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.