In the last few years, diesel engines for passenger cars and light commercial vehicles have achieved in Europe a level of success as never before and, to-· day, several manufacturers are looking at the US market for the close future. This has been possible thanks to the exploitation of turbocharging systems and highly flexible fuel injection equipment which have increased the specific power output of the engine, reducing in the same time the combustion noise and maintaining the high level of fuel efficiency typical of this kind of powertrain. On the other hand the complexity of diesel engines has increased exponentially, with the addition of several sensors and actuators necessary to monitor and control the correct working of the system. In order to achieve the increasing emission targets requested by European and US regulations, it is necessary not only to optimize the calibration of all the engine parameters, but also to ensure that the calibration setpoints are always achieved through an accurate control system. Moreover, it is necessary to be able to estimate how the tolerances errors of the engine sensors and actuators, inevitable in mass production equipment, will affect the robustness of the final engine outputs. In this thesis, experimental investigations have been executed on two different direct injection diesel engines, in order to characterize their emission behaviour and create some general prediction functions for the engine-out emission levels, based on mean value variables. These functions can be easily implemented in power plant engine models, in order to investigate the effect of deviations of single components on the whole engine system. Moreover, an explorative investigation has been carried out on the possible exploitation of in-cylinder pressure data for improving the control of the fuel injection and combustion, achieving an overall engine robustness advantage.