The simulated annealing algorithm with the Bessel method for curve fitting and the tensor product method for surface fitting is used to transform engine discrete experimental data into a form that enables these data to be incorporated in the optimization process. Optimum curves of the engine torque versus the engine rotational speed and the engine rotational speed versus the motorcycle speed for the fuel consumption and the carbon monoxide (CO) emission are obtained for a motorcycle with a continuously variable transmission (CVT). The engine rotational speed at which a motorcycle begins to move for the specific engine data is obtained. From design parameters, engine rotational speeds corresponding to the maximum and minimum CVT speed ratio change, the minimum fuel consumption and CO emission, and optimum design variables can be determined.