2.3 Implementing requirements using machine learning (ML)

Assurance objective: Provide a ML implementation that meets the defined safety requirements.

Contextual description: ML may be used as part of the implementation of any of the ‘SUDA’ functions, but in practice is most likely for Understanding and Deciding. Where ML is used as part of the implementation, it is necessary to ensure that the implementation satisfies the allocated safety requirements. Different types of machine learning technology may be adopted including neural networks, Bayesian networks and reinforcement learning, and the implications that technology choices may have on assurance must be considered.

This objective is achieved through the consideration of three sub-objectives as described below. These sub-objectives reflect the main elements of an ML process as shown in Figure 2 below.

Practical guidance:  Discussion of the capabilities and challenges associated with different ML technology that may affect adoption decisions for safety related RAS.

2.3 Approach to demonstration cross domain tag (new 2019)

Next sections:

Contact us

Assuring Autonomy International Programme
assuring-autonomy@york.ac.uk
+44 (0)1904 325345
Department of Computer Science, Deramore Lane, University of York, York YO10 5GH

Contact us

Assuring Autonomy International Programme
assuring-autonomy@york.ac.uk
+44 (0)1904 325345
Department of Computer Science, Deramore Lane, University of York, York YO10 5GH