DescriptionDiesel vehicles comprise a significant minority of the the vehicles in use today and contribute to a majority of the miles driven. With increasing focus on the environmental and health related risks of auto exhausts, regulations have been introduced to curtail emissions. Reliable sensors are crucial for monitoring compliance with regulations and optimizing performance of after-treatment systems. We show here work to develop computational models that can estimate exhaust gas concentrations emitted from diesel engines. We first introduce the problem in detail and briefly go over the experimental setup and sensor design, done by our collaborators at Los Alamos National Laboratory. We then look at a model based on the electrochemistry of gas-sensor interactions. This model is used to predict concentrations of gases in unknown two- and three-gas mixtures. We look at the performance of this model in predicting relative and absolute gas concentrations. The robustness of the model to reduction in sensor number and training data size is also analyzed. Next, we adapt a more flexible, machine learning model which is used to classify and quantify the gases in one- to four-gas mixtures. We look at the hyperparameters of this model to analyze the relative importance of each sensor and check how well the model performs on reduced sensor data. We also examine various modifications of this model and a compare results.