Evaluation of Non-Cooperative Sensors for Sense and Avoid in UAV Systems

05/09/2019

The number of civil and military applications for Unmanned Aircraft Vehicles (UAV) is increasing in the last years, such as firefighters, search-and-rescue missions or package delivery, among others. Due to their ability of performing a large variety of important tasks with higher manoeuvrability, longer endurance and less risk to human lives, small UAV are particularly suitable. But to carry out these tasks, it is mandatory to guarantee a safe performance and a correct integration into non-segregated airspace. Integrating unmanned aircraft into civil airspace requires the development and certification of systems for sensing and avoiding (SAA) other aircraft. In particular, non-cooperative Collision Detection and Resolution (CD&R) for UAV is considered as one of the major challenges to be addressed. The new project Enhanced Guidance, Navigation and Control for Autonomous Air Systems based on Deep Learning and Artificial Intelligence started by Boeing at the Center for Aerospace Research (CfAR) requires from definition of the SAA system and a rigorous analysis of it before the system can be developed and eventually certified for operational use. This paper will be focused on evaluating the capabilities of the non-cooperative sensors for a SAA system, reviewing the different sensors available and the data fusion techniques to merge the information provided by the different sources. Finally, algorithms for visual cameras image processing using machine and deep learning techniques will be developed and compared, with the aim to provide an effective obstacle detection capability.

Autores da comunidade :

• 

  P

  Pablo Arroyo Izquierdo

  ist191988

Orientadores desta instituição:

• 

  A

  Afzal Suleman

  ist13672

mechanical-engineering - Engenharia Mecânica

eng - Inglês

Instituto Superior Técnico