This project is concerned with the development of advanced deep neural network models for decision making under uncertainty, in particular for the application of energy-efficient navigation of electric vehicles. In this project, we will develop sophisticated methods to learn the parameters of the underlying mathematical models continuously as different problem instances are processed. This will help us to build models that do not rely on the availability of data a priori, and can improve the learning as more data is collected and attained. We will investigate the models for energy-efficient navigation of electric vehicles, where the goal is to learn and infer the path with optimal energy consumption. Our framework is generic and can include other objectives such as travel times as well.