Input-constrained adaptive GPC for simple industrial plant

One significant drawback of practical MPC schemes is the large computational burden, especially in adaptive and constrained situations. In this paper, previous work by the author towards a computationally efficient self-tuning Generalized Predictive Control scheme for low-order time delayed industrial plant is extended to include rate and amplitude constraints on the plant input. The paper discusses a scheme that has been optimized for real-time implementation on small, low-cost embedded processors, which avoids the need for a full QP procedure by optimizing over a short constraint horizon with a decomposed LQR terminal cost. Preliminary simulation and hardware-in-the-loop based experimental results indicate that the resulting scheme is indistinguishable from a full QP solution, and requires a fraction of the implementation overhead.