AI-Driven Runtime Monitoring of Energy Consumption in Autonomous Delivery Drones

Abstract

Last-mile package delivery has gained a lot of traction with the appearance of the COVID-19 pandemic. However, a wide range of issues, e.g. traffic congestion in urban areas and sparse transportation infrastructure in remote rural areas plague current package delivery processes. Autonomous delivery systems, e.g., autonomous delivery drones, are considered a viable alternative to improve the last-mile delivery process. In the development of an autonomous delivery system, part of the focus in system development lies with the design and development of the system’s safety-critical functions. Furthermore, these functions must be integrated into a smart and modular safety architecture. In this paper, we look at the energy management functionality of autonomous transport drones and develop a concept for the runtime monitoring of their energy consumption. Energy consumption prediction is a crucial element for preemptive energy management in autonomous delivery drones, which can be used to help prevent a failure of the delivery mission or even a drone crash. Our concept uses an AI-driven regression model in the form of a neural network, which dynamically predicts during system operation based on internal and external factors whether the drone is able to fulfill its mission or not. We evaluate the concept via simulation experiments and report on obtained results and lessons learned.