Products with embedded software have been around for many years. Most of these conventional embedded software products are built using rule-based control engineering approaches. The corresponding software engineering practices are mature and well understood. For autonomous systems, however, the conventional control engineering approaches are extended by modern artificial intelligence techniques, in particular machine learning. The corresponding product software engineering approaches are less well understood and need attention.

The goal of this workshop is to better understand the impact of incorporating machine learning algorithms in autonomous systems from the software engineering perspective and the implications on system properties such as quality, maintainability, scalability, robustness, safety, security, etc.

This workshop focuses on software engineering and software architecture approaches that achieve the typical software engineering goals for systems that are built using a combination of conventional embedded software development and AI.

The topics of interest include but are not limited to:

• Approaches to validation, verification, safety, reliability and their standardization
• Availability of shared data sets for training and validation and verification
• Data engineering and data management approaches for training and verification data and corresponding tool chains
• Labeling of training data as it relates to software engineering goals
• Simulation frameworks
• Methods to enrich data sets in order to achieve completeness and robustness, e.g., generation of synthetic data, decomposition and re-composition of natural data, synthetic modification of natural data
• Variant management of AI systems
• Distributed intelligent and cooperative controlling strategies
• Interface compatibility management, semantic service description for interoperability
• Data stream scaling in static networks
• Safety and security issues that are specific to the use of AI in autonomous systems