This paper describes an observer design method that estimates the exhaust manifold pressure in a diesel engine with limited sensor information. In the modern model-based control of air flows, information on the exhaust pressure is often necessary for control stability. However, it is not easy to measure the exhaust pressure with sensors due to the unfavorable environment for measurement and the high cost of sensors. Therefore, a robust observer, which is applicable to engines equipped with a dual-loop exhaust gas recirculation (EGR) system and a variable geometry turbine, is designed to estimate the exhaust pressure. The observer takes the form of a Luenberger-sliding mode observer in order to guarantee its asymptotic stability upon consideration of the model uncertainties. For research on state estimation and control of dual-loop EGR engines, information on the pressure states in the low pressure EGR loop must be provided. Hence, this paper also proposes an intuitive regression modeling method for the pressure states of the compressor inlet and turbine outlet with physical insights into the engine operation. Applying the regression models to the observer algorithm, the estimation performance of the exhaust pressure observer is verified experimentally in both steady state and transient conditions with a 6-L heavy-duty diesel engine.