This study deals with the problem of trajectory tracking of wheeled mobile robots (WMR's) under non-holonomic constraints and in the presence of model uncertainties. To solve this problem, the kinematic and dynamic models of a WMR are first derived by applying the recursive Gibbs–Appell method. Then, new kinematics- and dynamics-based multivariable controllers are analytically developed by using the predictive control approach. The control laws are optimally derived by minimizing a pointwise quadratic cost function for the predicted tracking errors of the WMR. The main feature of the obtained closed-form control laws is that online optimization is not needed for their implementation. The prediction time, as a free parameter in the control laws, makes it possible to achieve a compromise between tracking accuracy and implementable control inputs. Finally, the performance of the proposed controller is compared with that of a sliding mode controller, reported in the literature, through simulations of some trajectory tracking maneuvers.

This paper discusses the application of a constraint-based model predictive control (MPC) to mobile manipulation tracking problems. The problem has been formulated so as to guarantee offset-free tracking of piecewise constant references, with convergence and recursive feasibility guarantees. Since MPC inputs are recomputed at every control iteration, it is possible to deal with dynamic and unknown scenarios. A number of motion constraints can also be easily included: Acceleration, velocity and position constraints have been enforced, together with collision avoidance constraints for the mobile base and the arm and field-of-view constraints. Such constraints have been extended over the prediction horizon maintaining a linear-quadratic formulation of the problem. Navigation performance has been improved by devising an online algorithm that includes an additional goal to the problem, derived from the classical vortex field approach. Experimental validation shows the applicability of the proposed approach.

We consider feedback predictive control of a discrete nonhomogeneous Markov jump system with nonsymmetric constraints. The probability transition of the Markov chain is modelled as a time-varying polytope. An ellipsoid set is utilized to construct an invariant set in the predictive controller design. However, when the constraints are nonsymmetric, this method leads to results which are over conserved due to the geometric characteristics of the ellipsoid set. Thus, a polyhedral invariant set is applied to enlarge the initial feasible area. The results obtained are for a more general class of dynamical systems, and the feasibility region is significantly enlarged. A numerical example is presented to illustrate the advantage of the proposed method.

Concevoir une CN pour machine-outil selon une structure à architecture ouverte permet d'en améliorer la flexibilité en autorisant l'intégration de modules spécifiques. Dans ce contexte, l'article envisage en premier lieu l'ajout d'un module de commande d'axe basé sur une stratégie prédictive avancée. Cette stratégie incluant une anticipation en boucle fermée s'avère particulièrement performante en termes de suivi de consigne et de facilité d'implantation. Considérant le centre d'usinage MIKRON, des tests réalisés en simulation avec cette structure de commande, et comparés aux essais effectués avec une commande classique, valident l'apport de ce module. Une deuxième étape examine ensuite les perspectives apportées par l'intégration d'un module avancé de génération de consignes pour trajectoires d'usinage surfacique.

Common geared industrial robots call for force control methods with special properties such as good rejection of frictional disturbances, smoothness of corrective motions, and more. A new method is presented which meets these requirements and provides a high control bandwidth. In the manner of hybrid control, directions of a task frame can be selected to be force, impedance or position controlled. A joint-based inner position loop and a superimposed predictive force controller are used. Practical results include data from a robotic grinding facility. Here, the controller proved robustness and good performance under rough conditions.