Real-Time Evolutionary Algorithms for Constrained Predictive Control

In the last years, thanks to the great advancements in computing technology, Evolutionary Algorithms (EA) have been proposed with remarkable results as robust optimisation tools for the solution of complex real-time optimisation problems. In this chapter we review the most important results of our research studies concerning the design of EA-based schemes suitable for real-time optimisation problems for Nonlinear Model Based Predictive Control (MBPC). In the first part of the chapter it will be discussed some modifications of a standard EA in order to address some real-time implementation issues. The proposed extension concerns the adoption of a new realtime adaptive mutation range to generate smooth commands, and the adoption of an intermittent feedback to face the computational delay problem. It will be shown that the main advantage of the improved technique is that it allows an effective real-time implementation of the Evolutionary-MBPC with a limited computing power. The real-time feasibility of the proposed improved Evolutionary-MBPC will be demonstrated by showing the experimental results of the proposed method applied to the control of a laboratory flexible mechanical system characterized by fast dynamics and a very small structural damping. Later, the application of real-time EAs is exte nded to real-time motion planning problems for robotic systems with constraints either on input and state variables. Basing on a finite horizon prediction of the future evolution of the robot dynamics, the proposed EA-based device online preshapes the reference trajectory, minimizing a multi-objective cost function. The shaped reference is updated at discrete time intervals taking into account the full nonlinear robot dynamics, input and state constraints. A specialized Evolutionary Algorithm is employed as search tool for the online computation of a sub-optimal reference trajectory in the discretized space of the control alternatives. The effectiveness of the proposed method and the online computational burden are analyzed numerically in two significant robotic control problems.