ABSTRACT
In applying this approach to optimization problems it seems natural to base the error control on the given cost functional. In this particular case the cor responding dual solution is given in terms of the state variable, Lagrangian multiplier and control variable. Hence, the evaluation of the a posteriori error estimates does not require extra work for solving a dual problem and a posteri ori error estimation is almost for free. This observation leads to a particularly simple and efficient strategy for mesh adaptation in the finite element discretiza tion of optimal control problems. We provide the theoretical background for this approach and discuss some computational tests.
