An Autonomous Unmanned Aerial Vehicle System for Sensing and Tracking

Abstract

Cooperative unmanned aerial vehicle (UAV) teams can be used to accomplish many tasks such as surveillance and tracking without many of the normal risks associated with using manned vehicles. We present a software and hardware platform for a single user to interact with a UAV eet. A component-based software architecture built on a Linux platform is used to separate functionality. A bottom-to-top modularization is used in order to achieve the requirements of the system. This architecture makes adding functionality simple, while being exible enough to deal with hardware changes and robust enough to meet the demands of a mobile robotic system. A single user is able to interact with the system using a graphical user interface. Users are able to see the current status of aircraft and can input dierent types of tasks depending on the capabilities of the aircraft. Decentralized task allocation is done by the eet with onboard computing and the use of ad-hoc wireless communication. The tasks are allocated based on the user specied task requirements, aircraft capabilities, and a cost function which minimizes time to completion for each task. A gimbal is used to point any mounted sensors in the desired location. Video can be streamed to users interacting with the system and also recorded for later analysis. Vision-based algorithms allow for the probabilistic based tracking of pedestrians. Algorithms have been developed to give an optimized path for the UAV to follow based on a given state and the current target estimate probability distribution.