控制理论(社会学)
计算机科学
地平线
稳健性(进化)
跟踪(教育)
模型预测控制
控制工程
自适应控制
工程类
工作(物理)
数学优化
移动视界估计
时间范围
鲁棒控制
动态规划
弹道
跟踪系统
估计理论
作者
Qianyue Luo,Yukang Cui,James Lam
标识
DOI:10.1109/tase.2026.3703752
摘要
Conventional Model Predictive Control (MPC) often loses recursive feasibility when tasked with tracking aggressive or discontinuous dynamic reference trajectories. While existing Feasibility Governor (FG) frameworks effectively address this issue for steady-state setpoint regulation, they are often ill-suited for tracking continuously time-varying references, leading to conservative performance or tracking lag. To address this gap, this paper proposes a novel Dual-type Reference Governor (DRG) framework that seamlessly integrates with Tube-based MPC (TMPC). Unlike traditional governors, the DRG features a dual-mode architecture designed specifically for dynamic tracking: an Initial RG (IRG) handles large reference jumps to recover feasibility, while a Terminal RG (TRG) manages local trajectory refinement. Furthermore, to tackle the computational burden associated with robust control, a DRG-driven adaptive prediction horizon mechanism is introduced. This mechanism dynamically shortens the horizon during smooth tracking and extends it only when necessary for feasibility recovery. Theoretical analysis establishes the recursive feasibility and closed-loop stability of the proposed framework. Simulation studies and hardware experiments on a vehicle robot platform demonstrate that the DRG+TMPC significantly outperforms standard TMPC and conventional FG approaches in terms of tracking accuracy for dynamic targets and computational efficiency.
科研通智能强力驱动
Strongly Powered by AbleSci AI