Application and field validation of dual-reference-point fuzzy-gain Stanley path tracking for four-wheel independent...

High-throughput crop phenotyping requires accurate and smooth path tracking during inter-row travel and row transitions. For four-wheel-independent-drive and independent-steering agricultural robots, conventional single-reference geometric controllers usually use only one axle or one reference point for feedback, which can cause front-rear posture lag and oscillatory correction under disturbances. Here we propose a dual-reference-point fuzzy-gain Stanley controller that computes coordinated front and rear steering commands from the lateral and heading errors at both axles. The key distinction from conventional single-reference controllers is that the proposed strategy treats the front and rear axles as coordinated control objects, rather than using one axle or one reference point as the sole feedback source. An online fuzzy scheduler adaptively tunes the Stanley gain, and Ackermann relations map axle commands to individual wheel angles and speeds for four-wheel steering implementation. The method was evaluated using MATLAB/Simulink simulations and through pavement, potted-wheat, and rapeseed field tests. In straight-line simulation, settling time decreased from 11.03 s with conventional Stanley and 16.16 s with Pure Pursuit to 8.27 s, and further to 5.12 s with fuzzy gain scheduling, without noticeable overshoot. In field experiments, maximum/mean lateral deviation remained within 63/17 mm and maximum/mean heading error within 3.4°/0.9°. During row transitions, the corresponding initial deviations were within 35/29 mm and 2.9°/2.6°. These results show that the proposed controller enables accurate, smooth, and real-time tracking for practical crop phenotyping.