Abstract
The popularity of parallel robots has been steadily increasing in recent decades. This popularity has been strongly stimulated by the many advantages these robots offer over their serial counterparts, in some industrial applications requiring high acceleration and very good accuracy. However, in order to fully exploit their potential and make the most of their capabilities, a long way remains to be done. In addition to mechanical design, calibration, and structure optimization, the development of efficient control approaches plays a key role in improving the overall performance of these robots.Selective sorting consists in sorting and recovering waste according to its nature: metals, paper, glass, organic, etc, to facilitate its recycling. They are sorted by those who produce them, or by specialized organizations in sorting centers. The objective of this thesis is to study the use of parallel robots for Pick-and-Throw (P&T) applications in the selective and fast sorting of waste. The goal is to perform pick-and-throw tasks in a robust and fast way using a parallel manipulator (made available by Tecnalia under a collaborative research contract), demonstrating the interest and relevance of a P&T approach compared to a traditional Pick-and-Place (P&P) approach in the context of a selective waste sorting application.In this context, trajectory generation and control design are addressed in this thesis. On the one hand, motion planning for PKMs is not trivial. Different constraints such as kinematics, dynamics constraints, continuity, smoothness, etc., should be taken into account to generate a feasible and appropriate trajectory that meets the requirements of a specific application. On the other hand, from a control point of view, the control of PKMs is often considered in the literature as a challenging task due to their highly nonlinear dynamics, abundant uncertainties, parameter variation, and actuation redundancy.In this thesis, we aim to generate a fast and accurate P&T task using a parallel manipulator. Thus, we first propose a time-optimal P&T trajectory that significantly reduces the cycle time compared to the usual P&P technique. Real-time experiments have been conducted to validate the proposed P&T method, showing the relevance of this method with respect to the P&P process and to an existing P&T technique in the literature. Second, advanced robust control strategies have been proposed, which are extensions of (i) the standard RISE (Robust Integral of the Sign of the Error) feedback control, (ii) the DCAL (Desired Compensation Adaptive Law), and (iii) the model-free control (MFC). Lyapunov-based stability analysis is established for all the proposed controllers verifying the asymptotic convergence of the tracking errors. In order to validate the proposed controllers, numerical simulations are conducted on a parallel robot prototype, called T3KR. Several simulations are tested including robustness towards payload changes and robustness towards speed variations. The relevance of the proposed control schemes is proved through the improvement of the tracking errors at different dynamic operating conditions.