TY - JOUR
T1 - Kinematically feasible path planning and robust predefined performance formation control for MAGVs under mixed uncertainties
AU - Gao, Jia Wei
AU - Ge, Ming Feng
AU - Li, Yi Fan
AU - Wang, Yong
AU - Liu, Feng
N1 - Publisher Copyright: © 2026 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
PY - 2026/6/25
Y1 - 2026/6/25
N2 - Conventional path planning methods often suffer from low computational efficiency in unknown environments and frequently neglect the kinematic and dynamic feasibility of the generated paths. To address these challenges, this paper proposes a hierarchical planner-controller framework for multi-automated guided vehicles (MAGVs) that integrates optimal path planning with robust formation control subject to mixed uncertainties. Firstly, a novel hybrid Q-learning (HQL) algorithm is developed, which employs a distance-based initialization strategy, a Dubins-curve smoothing mechanism and the adaptive exploration rate to significantly accelerate convergence and ensure kinematic feasibility. Secondly, an adaptive fixed-time predefined performance controller (AFTPPC) is designed for the physical control layer. By introducing distributed fixed-time estimators and logarithmic error transformation mechanisms, it enables the controllers of all AGVs to ensure that the tracking error strictly converges within the specified range within a fixed time, regardless of the initial conditions or unmodeled dynamics. Simulation results demonstrate the framework’s superior robustness and efficiency, reducing the average path length by 29.4% and ensuring tracking errors converge within 0.5s, thereby maintaining high precision throughout the operation despite external disturbances.
AB - Conventional path planning methods often suffer from low computational efficiency in unknown environments and frequently neglect the kinematic and dynamic feasibility of the generated paths. To address these challenges, this paper proposes a hierarchical planner-controller framework for multi-automated guided vehicles (MAGVs) that integrates optimal path planning with robust formation control subject to mixed uncertainties. Firstly, a novel hybrid Q-learning (HQL) algorithm is developed, which employs a distance-based initialization strategy, a Dubins-curve smoothing mechanism and the adaptive exploration rate to significantly accelerate convergence and ensure kinematic feasibility. Secondly, an adaptive fixed-time predefined performance controller (AFTPPC) is designed for the physical control layer. By introducing distributed fixed-time estimators and logarithmic error transformation mechanisms, it enables the controllers of all AGVs to ensure that the tracking error strictly converges within the specified range within a fixed time, regardless of the initial conditions or unmodeled dynamics. Simulation results demonstrate the framework’s superior robustness and efficiency, reducing the average path length by 29.4% and ensuring tracking errors converge within 0.5s, thereby maintaining high precision throughout the operation despite external disturbances.
KW - Adaptive control
KW - Formation control
KW - Path planning
KW - Predefined performance control
KW - Reinforcement learning
UR - https://www.scopus.com/pages/publications/105035530720
UR - https://www.scopus.com/pages/publications/105035530720#tab=citedBy
U2 - 10.1016/j.eswa.2026.131904
DO - 10.1016/j.eswa.2026.131904
M3 - Article
SN - 0957-4174
VL - 317
JO - Expert Systems With Applications
JF - Expert Systems With Applications
M1 - 131904
ER -