Towards a Generalized SDK for a Programmable Drones-as-a-Service

The proliferation of Unmanned Aerial Vehicles (UAVs) in spatio-temporal applications has given rise to the need for Drone-as-a-Service (DaaS) platforms that support diverse tasks, ranging from urban air mobility to precision agriculture. However, existing DaaS platforms are often tightly-coupled to hardware-specific frameworks that lack programmability and flexibility for custom applications. For instance, using UAVs to help a Visually Impaired Person (VIP) navigate requires a suite of applications, such as obstacle avoidance, route planning, and real-time tracking. In this poster, we propose Ocularone SDK (O-SDK) that integrates on our prior Ocularone stack for VIP applications with Aerostack2, which generalizes the hardware abstraction. O-SDK offers an analytics layer as a novel contribution to enable developers to easily integrate Deep Neural Network (DNN)-based models, and interface swarms of UAVs for complex tasks like navigating the visually impaired. Our evaluations show that the minimal overheads of using O-SDK for real-time experiments using Tello drones and Jetson Orin Nano. This opens the potential to develop a domain-neutral, programmable and customizable DaaS Software Development Kit (SDK) in future.