Proportional-Derivative and Model-Based Controllers for Control of a Variable Mass Manipulator

Abstract

In the paper,numerical analysis of dynamics of a variable mass manipulator is presented. A revolute joints composed manipulator is considered. Payload of the gripper is considered as the only element characterized by unknown value of its mass (variable between subsequent operations). As in other cases of the revolute joints composed manipulators, its behaviour dependents significantly on the pose of the manipulator. When the manipulator is driven over a larger operating range, nonlinear terms can be observed in the equations, and linear controller does not ensure satisfactory its performance. Thus, nonlinear control technique is employed. To ensure such non-linear control, nonlinear models of the plant are introduced in the controller. Initially, there is an open-loop feedforward model-based controller used to enforce the manipulator to follow the required paths. Work of the proposed configuration is tested numerically. To deal with it, a numerical model of the manipulator is prepared and rules of multibody modelling are used for it. Two identical models (equations) are implemented in two separate blocks of the model. In the controller block, primary set of parameters of the robot structure is introduced. As the actual gaol is to observe effects of the incorrect identification of the payload masses, the plant (the controlled object) model may not be described by identical values of their masses in the performed tests. As a result, significant path errors are observed during simulations. To eliminate these errors, the controller model is enlarged. A feedback loop is accepted and a proportional-derivative controller is considered in this feedback part. Subsequently, collaboration between these feedforward and feedback blocks is tested. This collaboration runs well (error of the realised path is reduced). However, intensive work of the feedback controller is non-required in the considered application. Possibility of mass identification is tested. Performed tests proofed that signal of the proportional-derivative controller is a useful data base for the required identification process. A short run of a trial path is performed at beginning of the motion. PD signal obtained during such test motion are collected. They allow us to identify the mass of the payload and to improve data in the model-based controller. Thanks of it, the feedback control signal is reduced and realisation of the required path is acceptable.