Generalised predictive control with input constraints

Generalised predictive control (GPC), which has been shown to be effective for the self-tuning control of complex plants, is based on the minimisation of a long-range cost function. In previous work this has been achieved by an analytic solution which is not valid if there are constraints on the input control signals. A simple extension is proposed which caters for rate or amplitude limits, and simulations show the corresponding improvements in closed-loop performance that can be achieved. A common problem, in practice, is the effect of torque saturation in DC motor drives: results from a compliant link, where the control sample-rate is 60 Hz, show that the proposed method overcomes this and is also computationally acceptable. Bang-bang control, a limiting case of constrained inputs, is also treated within the GPC framework.