This paper describes an automatic approach for developing models for the xerographic process based on experimental data. A nonlinear system model defining a vector mapping between the actuators and performance metrics is constructed using multivariate adaptive regression splines (MARS). Multiobjective optimization techniques based on an underlying adaptive simulated annealing approach are discussed which are then used for determining the Pareto-optimal setpoints of xerographic actuators that optimize image quality metrics.The overall approach is applied to perform the automated setups of a high-end color machine, namely the Docucolor40. Experimental results show that high quality machine setups can be performed in an efficient and automated manner.