Evolutionary Optimization approach to tune a PID controlled robotic arm – Scilabtec 2015

By Taylor Newill, Noesis Solutions

ScilabTEC 2015 - Noesis from Scilab on Vimeo.

This work increases the stability and performance of a proportional-integral-derivative (PID) controller and provides a framework in which a PID controller can be easily tuned. PID controllers are widely used in almost all industrial control systems because of their well-known beneficial features. Tuning the controller in the proper manner is not simple. The tuning process is time consuming and the final result affects the entire system’s performance and plays a key role in the design optimization. In this work, a PID controlled robotic arm will be analyzed. The novelty of this work is the development of a new methodology for PID control tuning by coupling Xcos (and Scilab) within a process integration and design optimization platform such as Optimus. Optimus can easily integrate Scilab with all its capabilities as a computing environment for engineering and scientific applications. This is done by means of a “User Customizable Interface” (UCI) a wrapping layer able to analyze all Scilab scripts and access all Scilab parameters and results. The Xcos model is connected to Optimus, and Optimus automatically pilots Xcos to identify the optimal solutions of the robotic arm. Within the platform the use of evolutionary optimization algorithms (EAs) are applied. These algorithms are population based and can be efficiently parallelized. Moreover evolutionary algorithms can easily be used to solve multiobjective optimization problems. Most real-case optimization problems are multiple objectives in nature. This is true also for control problems where the objectives (e.g. rise time and stability) are often conflicting or competing. When approaching a problem as a multiobjective optimization problem, the designer is no longer searching for a single optimum, rather a compromise satisfying all the objectives and constraints. The set of compromise solutions is referred as the non-dominated points, or the set of Pareto points. Within this set, attempted improvement in one objective will result in degradation in one or more of the others. We solve the problem as a multiobjective optimization problem. The entire Pareto set is saved and all the results can be efficiently plotted and analyzed directly in the platform.

Co-signed paper by Silvia Poles & Barret Newill, Noesis Solutions