Nonlinear Predictive Control of Manipulator Arms

Abstract

The subject of the article are predictive control algorithms (of MPC type – Model Predictive Control) for rigid manipulator arms. MPC with a state-space model and with the latest disturbance and modeling error suppression technique was applied, which avoids dynamic disturbance modeling or resorting to additional disturbance cancellation techniques, such as SMC. First of all, the most computationally efficient MPC-NPL (Nonlinear Prediction and Linearization) algorithms are considered, in two versions: the first with constrained QP (Quadratic Programming) optimization and the second with explicit (analytical) optimization without constraints and satisfying a posteriori inequality constraints. For all considered algorithms, a comprehensive simulation analysis was carried out for a direct drive manipulator, with two kinds of disturbances: external and parametric. The obtained results were compared with those for the well-known CTC-PID algorithm (CTC – Computer Torque Control), showing better control quality with MPC algorithms. In addition, the influence of the length of the sampling period and of the computational delay on control quality was investigated, which is important for model-based algorithms with fast sampling.