The implantation of NO x sensors in diesel engines is necessary in order to track emissions at the engine exhaust for diagnosing and control of the after-treatment devices. However, the use of models is still necessary since the sensor outputs are delayed and filtered. The present paper deals with the problem of the NO x estimation in two parts, the first part deals with a controloriented model for the NO x estimation, while the second part presents data fusion of the model and sensor to improve the estimation, which is presented in the next. The use of models for the NO x estimation is an alternative but the drift and ageing are still an issue. In order to overcome this problem, the fusion of different signals can be made in a smart way by means of a Kalman filter. There exist different ways of presenting this fusion, from directly tracking the bias to updating the model parameters. According to that, different algorithms are proposed in this paper with the aim of correcting the model output. Furthermore, the estimation of the actual NO x , by preventing sensor delay and filtering, is also integrated in the algorithm, being a suitable strategy for combining NO x sensors and models in an on-board basis.