Exact boundary controllability of a hybrid system of elasticity (Q1110375)

The authors prove exact boundary controllability in arbitrarily short time for a system that models a flexible space structure consisting of an elastic mast carrying a rigid antenna at one end and clamped in a very massive space shuttle at the other end (the so-called SCOLE model). The physical problem is as follows: The shuttle may rotate in space but the antenna must point accurately in a specified direction and should thus be stabilized in the shortest time possible by applying forces and torques to it. The authors tackle a simplified version of this problem where all motions are planar. The system thus consists of an Euler-Bernoulli dynamical beam equation for the mast, taking into account inertial forces, and Newton's law for the position and rotation of the rigid antenna. As the beam and the antenna are rigidly linked together, the beam exerts forces and torques on the antenna and these two equations are coupled. The idea is to transform the ODEs for the rigid body into boundary operators for the beam, \(B_ 1\) and \(B_ 2\), in such a way that if \((B_ 1,B_ 2)w=(f_ 1,f_ 2)\), where \((f_ 1,f_ 2)\) are the force and torque applied to the rigid body, and if w satisfies the Euler-Bernoulli equation in the beam, then w is the actual deflection of the mast. At this point, the authors assume that the shuttle is fixed in space, so that no inertial forces are present and that it is legitimate to try to bring the system down to rest. The question is thus transformed into exact controllability of the Euler-Bernoulli equation with a clamping condition at one end and control through nonstandard boundary operators \(B_ 1\) and \(B_ 2\) at the other end. The authors first study in great detail the existence and smoothness of solutions in a semi-infinite beam with a clamped end by means of fundamental solutions. Then, given an initial condition for the finite beam (satisfying some technical assumptions of regularity), they extend it to an initial condition on the semi-infinite beam. They cut off the corresponding solution W(x,t) by a function \(\psi\) (t) with support in a strip \(0\leq t\leq \epsilon\) and \(\psi (t)=1\) for \(t<\epsilon /4\), \(\psi (t)=0\) for \(t>3\epsilon /4\). This function \(\psi\) (t) is chosen in the Gevrey class \(\gamma^ 2\). Then, they set \(w(x,t)=\psi (t)W(x,t)- Z(x,t)\), where Z is defined by an explicit series, whose convergence follows from Gevrey estimates, with support in time in [\(\epsilon\) /4,3\(\epsilon\) /4]. This Z is such that w is a solution which drives the semi-infinite beam at rest at \(t=\epsilon\). Finally, the values of \(B_ 1w\) and \(B_ 2w\) give constructively the desired smooth controllers, while w restricted to the finite beam is its corresponding deflection.