Whenever learning about Artificial Intelligence, you always end up reading about the famous chess match between IBM computer Deep Blue and Kasparoff 20 years ago. This was the first step on the stairway we are still climbing today. Back then, this first milestone showed that computers could perform amazingly well in tasks that were strictly linked to human cognitive capacities, such as playing chess, planning strategies and forecasting the opponent’s movements.
In the last decade artificial intelligence researchers powered by Google, Facebook and other big tech companies and universities world-wide, have developed new AI techniques and algorithms that have brought us to what some experts are already calling the fourth revolution.
Medicine, transport, cybersecurity, communications, finances and many other fields can benefit from Artificial Intelligence and this is not some kind of science fiction. The aerospace industry is also jumping on this bandwagon and GMV too is no slouch in this matter. In particular, GMV is leading the SAFETERM and AI-GNCAir projects for the European Defence Agency (EDA). These endeavors are two of the GMV’s cutting-edge assets in this field.
SAFETERM aims at using Artificial Intelligence (AI) to enhance the Flight Termination Systems (FTS) currently in use in Remotely Piloted Aircraft Systems (RPAS) operations. The main requirement for SAFETERM system is to increase the overall level of safety in handling emergency situations where a C2Link loss occurs. In this situation the Pilot in Command (PiC) will not be able to interact with the platform and, being unable to reach the predetermined Flight Termination Areas, will need to define a new suitable and safe landing area, through computer vision techniques.
Computer Vision (CV) focuses mainly on how computers can gain high-level understanding of their surroundings by means of digital images or videos. In plain words, Computer Vision techniques seek to help computers “see” the real world. As human beings, this task might seem an easy challenge, but vision perception in a dynamic physical world with an almost infinite capacity for variation is a highly complex affair that cannot be taken lightly.
Computer vision as a field is an intellectual frontier. Like any frontier, it is exciting and disorganized, and there is often no reliable authority to appeal to. Many useful ideas have no theoretical grounding, and some theories are useless in practice; developed areas are widely scattered, and often one looks completely inaccessible from the other.
— Page xvii, Computer Vision: A Modern Approach, 2002.
At this point, it is important to bear in mind that computer vision is not the same as image processing. The latter is more related to creating a new image from an existing one, by enhancing or simplifying the data in some way. One would say that image processing is the first step to preparing the data input for the Computer Vision workflow.
Amongst all the possible CV applications, SAFETERM is based on area recognition: what areas are there in the image and where are they. In this project, another of EDA’s goal is to weigh up the challenges of using AI technology in aviation. A real avionics hardware and software may be developed following aviation standards that evaluate the certification items and milestones for such AI-based embedded systems.
The AI-GNCAir project involves another kind of technique; it focuses self-localization data-fusion, coming from air-vehicle sensors. “Autonomous and automated decision-making techniques for manned & unmanned system” and “multi robot control and cooperation” are the aims of the project, where GNC systems shall be enhanced by using Artificial Intelligence techniques for data and information management.
In this field, security aspects are of great importance. Confidentiality, not allowing external interferences; integrity, safeguarding the accuracy of data as it moves through the workflows and availability, as the dataflow must never be interrupted. AI algorithms must be able to recognize signal interferences, incorrect readings from sensors, or even predict data that might be missing due to any of the abovementioned factors.
Some of the fields where AI-GNCAir focuses on are robust data recordings, efficient data-fusion protocols, data-fusion computational complexity management or dynamic sensor selection for continuous data availability.
Artificial intelligence and machine learning are generic terms for a wide variety of data processing, control and optimization techniques applicable to almost any industry or system. Air-vehicles can benefit from these cutting-edge technologies which will lead to further autonomy, safety and will allow human operators to provide more high-level input and supervisory control.
Authors: Javier Ferrero and Eugenio Sillero.
Las opiniones vertidas por el autor son enteramente suyas y no siempre representan la opinión de GMV
The author’s views are entirely his own and may not reflect the views of GMV