Contribution à la modélisation dynamique, l'identification et la synthèse de lois de commande adaptées aux axes flexibles d'un robot industriel.

摘要

Anthropomorphic robots are widely used in many fields of industry to carry out repetitive tasks such as pick and place, welding, assembling, and so on. Due to their flexibility and ability to perform complex tasks in a large workspace, industrial robots are finding their way to realize continuous operations. Then, high level pose accuracy is required to achieve a good path tracking. Unfortunately these systems were designed to have a good repeatability but not a good accuracy. The dynamics of these manipulators is subject to many sources of inaccuracy. Indeed, friction, kinematic errors and joint flexibilities may be the seat of deformation and vibration which degrade the position performance. These physical phenomena are even more difficult to manage even only a subset of states of the system is measured by motor encoders. Hence, the structure of current industrial control does not act directly on these phenomena. Nevertheless, there is a growing interest from industry for an improved path tracking accuracy with standard robots controllers. A state of the art highlights a lack of works considering the development of expectations adapted to the axes of an industrial robot and incorporating deformation phenomena. The approach proposed in this PhD. Thesis is meant to be an alternative to such techniques by proposing a methodology based on exploitation of detailed physical modeling and associated to experimental identification methods. This model incorporates the main highlighted physical phenomena. It is then exploited to obtain adapted control structures and tuning methods allowing enhancing the system's performance. It is integrated in our trajectory planner in order to realize a compensation scheme of joint errors. Thus, we introduce a new non-asymptotic estimation method applied in robotics, to on-line estimate the vibration parameters and to update operated models. Experimental results show that the proposed methodology leads to an improved motion control of the point-tool.