This thesis investigates mean-value modelling and robust control of the airpath of a diesel engine equipped with variable geometry turbocharger (VGT) and exhaust gas recirculation (EGR). Both the EGR and the VGT are driven by the exhaust gas and render the control problem inherently multivariable. In order to allow for model-based control design, a mean-value model of the airpath of the engine is derived with a focus on the parameterisation of the turbocharger. The effect of heat transfer via the turbocharger housing on the efficiencies derived from temperature measurements is identified as being very significant at low speed and load points regularly met on emission drive cycles. A physics-based parameterisation for the turbine efficiency map, which is the most difficult to model, is suggested. Based on experimentally obtained frequency responses, the parameters which are most uncertain in the model, i. e. compressor and turbine efficiency as well as the effective area of the EGR valve, are identified to have the biggest effect on the frequency responses. Different parameterisations of these uncertainties are then used for extended Je00 loopshaping design at a fixed engine operating point. Applying ?, analysis tools, it turns out that the application tailored uncertainties yield a better controller performance, which is confirmed by experimental data. In order to extend the controller operating regime, the nonlinear model is simplified and converted to linear parameter-varying (LPV) form . A robustly gain scheduled LPV controller is synthesised for this model using a gridding approach for the intake manifold pressure as scheduling variable. The designed controller is implemented on the engine in real-time. The experimental results are very promising and indicate that the quasi-LPV model captures the significant nonlinearities and dynamics of the plant.