مدل سازی پارامتر های متغیر خطی برای ترکیب برنامه ریزی اکتساب کنترل مقاوم ربات های صنعتی مشترک قابل انعطاف

Industrial robot features a multivariable, nonlinear and coupled dynamics property, which is always the inevitable topic for control engineer to face during the design of motion controller to achieve desired performance. In the past half century, motion control of industrial robot has gone through a sustainable development from no model involved to model based, from rigid body model to flexible joint model, and even more delicate models of higher order with flexibility in non-drive-train components are also introduced lately. Involvement of gradually refined dynamic model into control design, as well as manipulator design, has become a dominating approach for robot manufacturer to pursue competitive performance and keep technological leadership. Focusing on dynamic modeling, this paper introduces the formulation and approximation of a novel linear parameter-varying (LPV) model for the three main axes of typical six degrees-of-freedom (DOF) elbow type robot, which converts the strong nonlinear system into a quasi linear one globally dependent on certain scheduling parameters. This modeling method, as the prerequisite step for gain-scheduling robust control synthesis, paves the way for further step towards the implementation of LPV gain-scheduling modeling and control techniques for full robot in the next step.

variation can be modeled in the LPV formulation. In this paper, the LPV model for three main axes of 6- DOF robot is derived and ends up with six initial scheduling parameters. Then parameter set mapping based on PCA method is used to find out two new scheduling parameters so that computation cost and conservatism of the controller are much reduced. With such an LPV model, corresponding linear controllers at the vertices can be synthesized in the standard robust control design framework and LPV gain-scheduling controller can be obtained in the polytopic form dependent on the same scheduling parameter vector as the model accordingly