Due to increased environmental and economic considerations, energy-efficient vehicles such as Hybrid Electric Vehicles (HEVs) have received great attention from the general public as well as automotive research community. HEVs achieve better fuel economy than conventional vehicles by employing two different power sources: a mechanical engine and an electrical motor. By controlling the two power sources optimally based on the current system state under different driving conditions, they can achieve much improved fuel economy compared to conventional vehicles powered by internal combustion engines only. These power sources have conventionally been controlled by rule-based or optimization-based control algorithms. Rule-based algorithms utilize a well-defined and easy-to-understand control logic while optimization-based control algorithms employ a mathematical function that is minimized by a controller during the vehicle operation. Besides these two conventional approaches, recent advancements in machine learning offer new opportunities in optimal control of multiple power sources in unprecedent ways. Therefore, in order to investigate benefits offered by the new machine learning-based powertrain control paradigm, different machine learning approaches are studied for a HEV in this dissertation.