Extending the Range of Delivery Drones by Exploratory Learning of Energy Models

Abstract

Delivery drones have a fairly short range due to their limited battery life. We propose new exploration strategies to generate paths for a drone to reach its destination while learning about the energy consumption on each edge on its path so as to optimize its range in future missions. As the energy consumption mostly depends on the payload, the wind direction, and the wind speed, we developed an energy model to estimate the energy consumption based on these factors. We evaluated our exploration strategies for learning the energy model in order to identify the set of all reachable destinations. We found that adding a small amount of perturbation to encourage exploration can increase the learning rate.