Software Engineering for Intelligent and Autonomous Systems: Report from the GI Dagstuhl Seminar 18343
This work tackles the problem of system reliability and adaptability in critical domains like security and safety, but it is incremental as it consolidates existing research from multiple communities.
The report addresses the challenge of software systems operating in uncertain environments with evolving requirements and unexpected failures, which can lead to security, safety, or revenue issues, by summarizing research on using closed-loop control for intelligent and autonomous systems to enable dynamic adaptation without human involvement.
Software systems are increasingly used in application domains characterised by uncertain environments, evolving requirements and unexpected failures; sudden system malfunctioning raises serious issues of security, safety, loss of comfort or revenue. During operation, these systems will likely need to deal with several unpredictable situations including variations in system performance, sudden changes in system workload and component failures. These situations can cause deviation from the desired system behaviour and require dynamic adaptation of the system behaviour, parameters or architecture. Through using closed-loop control, typically realized with software, intelligent and autonomous software systems can dynamically adapt themselves, without any or with limited human involvement, by identifying abnormal situations, analysing alternative adaptation options, and finally, self-adapting to a suitable new configuration. This report summarises the research carried out during SEfIAS GI Dagstuhl seminar which provided a forum for strengthening interaction and collaboration for early-career researchers and practitioners from the research communities of SEAMS, ICAC/ICCAC, SASO, Self-Aware Computing and AAMAS.