Disturbance Compensation for Nonlinear Model Predictive Control Using Adaptive EKF with Disturbance Estimation

Nonlinear model predictive control (NMPC) is an advanced control method capable of achieving optimal motion control while considering system nonlinearities and constraints. Its performance relies heavily on accurate dynamic models and precise state measurements. In real-world applications, constructing an accurate model is difficult because of the presence of disturbances and noise, and directly measuring all system states is typically impractical. Therefore, state estimation that considers disturbances and noise is essential. However, disturbances and noise can degrade state estimation accuracy, thereby degrading NMPC tracking performance. To overcome these issues, this study proposes an offset-free NMPC framework that integrates NMPC with an adaptive extended Kalman filter (AEKF) featuring disturbance estimation. This integration improves state estimation by accounting for external disturbances, and the estimated disturbances can be used to compensate for both external disturbances and internal model errors. However, enhancing the disturbance estimation of observers tends to reduce the ability to filter out noise. To mitigate this trade-off, a variable law is introduced into the AEKF to adaptively improve its noise suppression ability. The proposed framework effectively balances disturbance compensation and noise filtering, thereby improving the overall control performance. Numerical simulations verified the effectiveness of the proposed framework.