Surrogate‐Assisted Nonlinear Model Predictive Control for a Lambda Robot

ABSTRACT This work presents a nonlinear model predictive control (NMPC) framework for a lambda‐shaped parallel robot that incorporates data‐driven surrogates to facilitate control design. An inverse surrogate model is used to provide an efficient initialization of the control inputs for NMPC. The presented controller enables anticipation of the system behavior, enforces constraints, and optimizes the control performance over a finite prediction horizon. The inverse surrogate model is trained using an appropriately generated data set. The resulting initialization, therefore, is already close to the optimal solution, which significantly accelerates the optimization process in the NMPC framework. Simulation results demonstrate that the proposed method achieves high accuracy and robust tracking performance in the presence of parametric uncertainty. Furthermore, comparative studies with a sliding mode controller (SMC) indicate that the proposed method exhibits performance characteristics that are comparable to those of SMC, where each approach shows distinct advantages depending on the specific operating conditions.