One of the most important problems which face the automotive industry in that of reducing fuel consumption, in view of the increasing cost of energy and the need to save oil for more sophisticated uses, while keeping pollutants emissions as low as possible, in view of the resulting health and environmental problems. Reduction of fuel consumption can be obtained working on different features of the car, including type of engine, aerodynamical design, used materials and optimal operations of the engine. In this work we shall be concerned with the last approach, which can result in a 5-20% saving on fuel, depending on which parameters are optimized. Mathematically, the problem can be formulated as follows: let s(t) be the value of a vector of state parameters (typically speed and power) at time t; then one has to choose optimal values of control parameters u(t), typically spark advance angle, air-fuel ratio and possibly transmission ratio, in such a way that the total fuel consumption on a certain cycle (EPA or European cycle) be minimized subject to constraints, of legal type, on the total amount of certain emitted pollutants (typically CO, NO X , HC) during the cycle and of course to drivebility constraints. The resulting problem is formally an optimal control problem, which is in practice discretized and thus reduced to a mathematical programming problem, of the nonlinear type and of moderately large dimensions (the number of independent variables is in practice about fifty, with about twice that number of constraints). One of the main difficulties in actually solving the above problem is that the form of the functions describing the rate of fuel consumption in terms of s(t) and u(t) is not available analytically and the same applies for the function describing the rate of pollutants emissions. We review the different approaches taken in the literature to estimate such functions, in particular the approach based on two-dimensional splines which has been succesfully adopted by us. Then we describe the various techniques considered in the literature for actually soving the resulting mathematical programming problem, including simplifications to reduce its dimensionality. We describe the approach that is under way in Bergamo and which also allows sensitivity analysis. Finally we conclude with some considerations on the effect of car aging on the optimal control parameters and some ideas on how to update both the model and the optimal parameters.