Task-space regulation of cooperative manipulators (Q1977046)

This paper is concerned with the position and orientation stabilization problem in the task space of a pair of non-redundant 6 d.o.f. rigid manipulators holding tightly a common rigid object. The control task has been described in the language of unit quaternions. The dynamics of the system have the form \[ \begin{cases} M(q)\ddot q+C(q,\dot q)\dot q+F\dot q+g(p)=\tau-J^T(q)h\\ M_{\text{ext}}\dot\nu_{\text{ext}}+C_{\text{ext}}\nu_{\text{ext}}+g_{\text{ext} }=h_{\text{ext}},\end{cases} \tag{*} \] where \(q\in\mathbb{R}^{12}\) denotes the joint position vector of the manipulators, \(h\in\mathbb{R}^{12}\) is the vector of contact forces exerted by end effectors, and the meaning of other terms is standard. The second line equations refer to the dynamics of the object. Using the pseudoinverse of the grasp matrix \(W\) the contact forces are expressed by external and internal forces acting on the object as \[ h=W^+h_{\text{ext}}+V h_{\text{int}}. \] The main result of the paper consists in proposing the following control algorithm \[ \tau=J^T(W^+W+V\Sigma V^+)K_pe-K_d\dot q+g+J^T W^+g_{\text{ext}}\tag{**} \] which is a variant of the PD controller with gravity compensation. \(\Sigma\) is a diagonal matrix able to suppress the influence of internal forces. Asymptotic stability of the closed loop system (*) and (**) has been derived from the La Salle invariance principle. Further modifications of (**) have been set out including adding an internal force feedback and introducing an estimate of the gravity term.