An energy management system for a vehicle is disclosed. The vehicle includes one or more power sources configured to provide power to one or more recipients. The system includes a controller configured to determine an arbitration vector based at least partially on a state vector and an initial transformation function. The arbitration vector is determined as one or more points for which the initial transformation function attains a maximum value. The controller is configured to determine a current reward based on the arbitration vector and the state vector, the current reward being configured to minimize energy loss in the power sources. The controller is configured to determine an updated transformation function based at least partially on the initial transformation function and a total reward. The controller is configured to arbitrate a power distribution based in part on the updated arbitration vector.