ABSTRACT
The configuration selection methods are equally applicable to other robot calibration techniques in which linear transformations relating pose measurements to the unknown kinematic parameters are available. Qualitatively, optimal selection of robot configurations can be stated as the problem of determining a set of robot measurement configurations within the reachable robot joint space so that the effect of measurement noise on the estimation of robot kinematic parameters is minimized. Simulated annealing (SA) as a stochastic search approach has been introduced to allow the iterative solution to escape local minima by occasionally accepting bad points. It has been successfully applied to a variety of optimization problems. The generation of "good" candidate states is critical for having a good convergence rate of the SA algorithm. A cooling schedule has to be carefully devised in order to guarantee the convergence of the algorithm.
