A recently proposed adaptive model predictive control algorithm is implemented in real time and its performance is verified experimentally on a quad-tank testbed. The algorithm relies on an iterative set membership identification procedure in order to provide a set of possible plant models at each time step. This set is then exploited by the model predictive controller in order to robustly enforce output constraints. The experimental results show that the algorithm provides good reference tracking performance while robustly satisfying output constraints for different operating conditions.
I. INTRODUCTIONModel predictive control (MPC) is a control strategy in which the input to be applied to the plant is calculated by solving an optimization problem at each time step, which allows one to enforce constraints on the plant's inputs and outputs. When the system to be controlled is uncertain, an attractive idea is to combine MPC with an on-line identification procedure, in order to gradually improve performance over time, as more data is collected from the system, while at the same time still satisfying the input and output constraints.
This idea has been explored by several researchers ([1], [2], [3], [4]).However, in the literature there is a lack of contributions providing evidence that such adaptive MPC approaches can be implemented successfully on a real-world system that is subject to constraints. Recently an adaptive MPC algorithm for open-loop stable, linear MIMO systems, able to deal with both input and hard output constraints as well as measurement noise and output disturbances, has been proposed ([5], [6]). The controller relies on a recursive set membership (SM) identification algorithm and is suitable for on-line implementation as it requires only the solution of standard convex optimization problems. In this paper, after briefly summarizing the approach, we present an experimental verification of this algorithm on a quad-tank testbed. The latter is a MIMO system that consists of four water tanks and two pumps [7]. The operating point of the system can be adjusted such that the dynamics exhibit linear behavior with either minimum or non-minimum phase. In addition, there are operating points in which the system dynamics are considerably nonlinear. These aspects make such a system a challenging testbed for any adaptive control scheme.