With increasing numbers of electric and hybrid vehicles on the road, transportation presents a unique opportunity to leverage data-driven intelligence to realize large scale impact in energy use and emissions. Energy management in these vehicles is highly sensitive to upcoming power load on the vehicle, which is not considered in conventional reactive policies calculated at design time. Advancements in cheap sensing and computation have enabled on-board upcoming load predictions which can be used to optimize energy management. In this work, we propose and evaluate a novel, real-time optimization strategy that leverages predictions from prior data in a simulated hybrid battery-supercapacitor energy management task. We demonstrate a complete adaptive system that improves over the lifetime of the vehicle as more data is collected and prediction accuracy improves. Using thousands of miles of real-world data collected from both petrol and electric vehicles, we evaluate the performance of our optimization strategy with respect to our cost function. The system achieves performance within 10% of the optimal upper bound calculated using a priori knowledge of the upcoming loads. This performance implies improved battery thermal stability, efficiency, and longevity. Our strategy can be applied to optimize energy use in gas-electric hybrids, battery cooling in electric vehicles, and many other load-sensitive tasks in transportation.

Human driven systems present a unique optimization challenge for robot control. Generally, operators of these systems behave rationally given environmental factors and desired goals. However, information available to subsystem controllers is often incomplete, and the operator becomes more difficult to model without this input information. In this work we present a data-driven, nonparametric model to capture both expectation and uncertainty of the upcoming duty for a subsystem controller. This model is a modified k-nearest neighbor regressor used to generate weighted samples from a distribution of upcoming duty, which are then exploited to generate an optimal control. We test the model on a simulated heterogeneous energy pack manager in an Electric Vehicle operated by a human driver. For this domain, upcoming load on the energy pack strongly affects the optimal use and charging strategy of the pack. Given incomplete information, there is a natural uncertainty in upcoming duty due to traffic, destination, signage, and other factors. We test against a dataset of real driving data gathered from volunteers, and compare the results other models and the optimal upper bound.