Machine Learning for Avionics

Presented By Dr. Pavel Paces


In this course we will explore the term machine learning and define algorithms to be generally considered as machine learning. The course is built around use cases where machine learning can provide advantage in form of time and cost savings. We are going to link the use of machine learning to existing algorithms used for system diagnostics which include signal processing algorithms, feature extraction and classification methods. The tutorial will begin with Signal to Noise Ratio, variance, Standard Deviation and FFT which can be used for unsupervised, supervised and reinforcement learning where such as regression, k-nearest neighbors and other algorithms are used. The tutorial will also introduce the basics of the neural networks, their design and pros and cons with explanation why certification authorities do not accept systems using neural networks for safety critical applications. The tutorial will be concluded by a use case utilizing machine learning with data classification algorithms for automatic recurrent testing of avionics software modifications.

Pavel Paces is currently member of Artificial Intelligence Center and Department of Aerospace Technologies at Czech Technical University in Prague, Czech Republic. He graduated from Electrical Engineering in 2005 and got his Ph.D. in Aerospace Engineering in 2011 at the same university. Pavel has past experience with aerospace sensors development, flight simulators certification and business development.