Ph.D. received on:  8/7/1997

E-mail: caccavale@disna.dis.unina.it

Tutor: Prof. Lorenzo Sciavicco, Università degli Studi di Napoli Federico II 

 ___________________________________________________________________________________________________________

Supervised Switching Control   ___________________________________________________________________________________________________________

Advisor:

Prof.  Lorenzo Sciavicco, Università degli Studi di Napoli Federico II 

Summary:

It has been recognized that the use of multiple manipulators in lieu of single manipulators has many advantages in terms of loading capabilities and manipulation dexterity. In this thesis the problem of modeling and controlling a system of two cooperative manipulators tightly grasping a common rigid object is addressed. The system can be seen as a closed chain of rigid bodies (i.e., the manipulators links and the held object) characterized by a set of constraints on the motion variables. First, the relevant variables for the kinetostatic description of a closed-chain of the whole system are recognized. To the purpose, two different formulations are analyzed, and the relevant mappings between generalized forces and velocities are derived. Then, the dynamic model of the system composed by the two manipulators and the held object is derived together with the expression of the closed-chain constraints.
Based on the results of the modeling analysis, the problem of the force/motion regulation is addressed. When a cooperative system is considered, both the absolute motion of the held object and the internal forces (i.e., mechanical stresses) acting on the object are to be controlled. To the purpose, a kinematic control strategy is devised, in which the joint-space set points are generated by solving the inverse kinematics of the cooperative system. If the joint-space motion is controlled via a PD regulator with gravity compensation, internal forces at steady state may arise when the joint references are not compatible with the closed-chain constraints. To overcome this problem, kinetostatic filtering of the control action is introduced, such that the components of the control action causing internal forces at steady state are filtered out from the control law. Also, direct internal force feedback can be added to improve the transient performance and the robustness of the proposed regulator.
Also, a PD regulator designed in the cooperative task-space is proposed, where the set-points for the regulator are directly specified in terms of absolute and relative motion of the system. Kinetostatic filtering is added to ensure null internal forces at steady state. In order to avoid problems coming from representation singularities, tracking errors in the control law are computed through unit quaternions. The stability of the equilibrium for the proposed control laws is demonstrated via invariant set theorems.
 Finally, the proposed control laws are experimentally tested on dual-arm system composed by two industrial manipulators with open control architecture. The obtained results have demonstrated the feasibility of the approach based on the kinetostatic filtering of the control action even on an industrial setup.
 

 
 _______________________________________