Path tracking accuracy and stability are crucial for Automated Guided Vehicles (AGVs) in intelligent manufacturing. Traditional Pure Pursuit (PP) methods involve an inherent trade-off: small look-ahead distances provide quick response but cause oscillations, whereas large distances improve smoothness at the expense of corner-cutting. To overcome this, we propose an A daptive M ulti-Scale F usion P ure P ursuit (AMSF-PP) method. It integrates adaptive Loess path preprocessing for noise suppression, a multi-step predictive controller for rapid local response, and a traditional PP controller for global guidance, combined via a real-time weighting mechanism. Simulation results show that AMSF-PP reduces average lateral error by up to 59% compared to conventional PP. Notably, in real-vehicle experiments on a resource-constrained industrial PC, AMSF-PP achieves higher tracking precision than Model Predictive Control (MPC) while being approximately 21 times more computationally efficient (1.15 ms vs 24.50 ms). These findings establish AMSF-PP as a lightweight, efficient, and robust solution for high-precision industrial path tracking.
Fu et al. (Fri,) studied this question.