Improved evolutionary predictive controllers for real-time application

In the last few years, nonlinear model-based predictive control (MBPC) has emerged as a technique to approach difficult multivariable control problems in the presence of constraints either on input or on state variables.The implementation of nonlinear MBPC requires the online solution of a nonconvex, constrained nonlinear optimization problem. Many conventional iterative and discrete search algorithms have been proposed as optimization tools, but very often problems related to local minima and computational complexity limit their efficiency. Some recent simulation studies have clearly shown the high potentiality of evolutionary algorithms for global optimization in nonlinear MBPC. In this article we propose a significant improvement of the basic evolutionary MBPC with the aim of addressing some real-time implementation issues. The proposed extension regards the insertion of a new online adaptive mutation range to generate smooth commands, and the adoption of an intermittent feedback to face the computational delay problem. The main advantage of the improved technique is that it allows an effective real-time implementation of the MBPC with a limited computing power. The real-time feasibility of the improved evolutionary MBPC has been experimentally demonstrated by applying the proposed method to the control of a laboratory flexible mechanical system characterized by fast dynamics and a very small structural damping.