2021/10/01 by Lukas Brunke, Brunke, Lukas, Siqi Zhou +3
Engineering · Medicine · #Advanced Control Systems Optimization #Advanced MRI Techniques and Applications #FOS: Computer and information sciences #FOS: Electrical engineering #Iterative Learning Control Systems #Robotics (cs.RO) #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2110.00542
openalex publication_date 2021/10/01 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
In this work we address the problem of performing a repetitive task when we\nhave uncertain observations and dynamics. We formulate this problem as an\niterative infinite horizon optimal control problem with output feedback.\nPreviously, this problem was solved for linear time-invariant (LTI) system for\nthe case when noisy full-state measurements are available using a robust\niterative learning control framework, which we refer to as robust\nlearning-based model predictive control (RL-MPC). However, this work does not\napply to the case when only noisy observations of part of the state are\navailable. This limits the applicability of current approaches in practice:\nFirst, in practical applications we typically do not have access to the full\nstate. Second, uncertainties in the observations, when not accounted for, can\nlead to instability and constraint violations. To overcome these limitations,\nwe propose a combination of RL-MPC with robust output feedback model predictive\ncontrol, named robust learning-based output feedback model predictive control\n(RLO-MPC). We show recursive feasibility and stability, and prove theoretical\nguarantees on the performance over iterations. We validate the proposed\napproach with a numerical example in simulation and a quadrotor stabilization\ntask in experiments.\n