To describe formal learning techniques for non-linear and adaptive systems using certifiable optimisation techniques in on-board avionics, and to analyse their impact on avionics hardware (processing and link functions) on verification and on validation.

 

There are problems that cannot be solved by traditional algorithmic solutions, especially in the Control domain. They require non-linear adaptive algorithms. Neural Network is one of them.

 

The term "Artificial Intelligence" is used intensively and is subject to many interpretations. The need is to build a system that is able to learn, adapt, and perform optimally in a non-linear environment using non-linear behaviour for space applications (such as Visual Space Navigation). The activity will first describe the field of avionics problems that "AI" can solve, and the related technology solutions, in order to produce a reference document defining the language used in the avionics community and its applicability for Space.

 

The activity will then focus on a use case, which is a Visual-Navigation system based on Relative Navigation with respect to celestial bodies, in particular looking at Image Processing in very low light conditions, for which Neural network is expected to be a solution among others.

 

The activity will implement, for this use case, a demonstrator based on software and hardware, with full characterization of the potential hardware solutions and their performance. Hardware software co-design will be considered.

 

The activity will finally describe the verification and validation process of such technique, including the definition of the related metrics, the training and validation sets, and the statistical evaluation of the potential errors. It will address the "Formal Learning"and the "Certifiable Optimisation Techniques"

 

The activity include the following tasks:

- taxonomy on "artificial intelligence" application and techniques focussed on on-board avionics

- specification and design of the use case

- specification and design of the software and hardware solution

- description of the validation process