Feedback Control Of Industrial Robots By Means Of Stochastic Optimization Techniques

Abstract

AbstractThe standard engineering approach in control of industrial robots assumes all model parameters are known exactly and therefore can be modeled by their nominal values. Due to stochastic variations of the material, manufacturing errors, modeling errors and most important stochastic variations of the workspace environment (e.g. stochastic payload mass) this assumption does not hold. Hence, an approach that models these parameters by means of random variables was proposed, cf. [1]. Solving this substitute optimal control problem by means of direct optimal control discretization techniques allows the numerical computation of approximate optimal stochastic reference trajectories for the on‐line control process. These optimal reference trajectories are ‐ according to the chosen probability distribution ‐ valid for a variety of parameter configurations and are more stable toward parameter uncertainties. However, due to the stochastic parameters in the robots dynamic model, in the on‐line control process still deviations from the reference trajectories appear. Hence, a method for the construction of feedback controllers based on feedback linearization is presented.