CONtinuous engineering and TRustworthy operation of Ai-enabled SysTems

Project description

Our visions of future Cyber-Physical Systems (CPS) are enabled by a trinity of technological characteristics – automation, interconnection, and AI. Inherent to these traits are unknowns and uncertainties which make it very difficult to ensure trustworthy (robust, safe, secure, transparent, and accountable) operation, because established methods and techniques rely on a complete understanding of a system and its context already at design and development time. Without being trustworthy, systems can hardly become an economic success and market introduction might even be completely impossible, especially in cases where critical requirements (e.g. safety) cannot be guaranteed. Already today, we can see disruptive changes across industries in adopting new system paradigms and it is to be expected that in the global competition market shares will be distributed anew. The US and China, as prosperous environments for countless start-ups and AI powerhouses, presently seem to lead the race. Europe consequently needs to strengthen its industries and push for promising USPs. Systems that are trustworthy, while at the same time providing outstanding performance, would clearly provide such an USP and can hence be key to succeed in this competition – Trustworthy systems made in Europe. To address this important challenge, CONTRAST is setting out to develop a runtime assurance approach for ensuring trustworthiness of complex CPS and Cyber-Physical Systems-of-Systems (CPSoS) with AI and ML components. In particular, the approach shall enable a continuous optimization of system performance while guaranteeing key properties of trustworthiness. To this end, systems shall be enabled to: (1) Continuously assess assumptions regarding a systems context and actual operational status and correspondingly determine its current capabilities w.r.t. provided functions and qualities. This particularly includes the capabilities of AI components, which will be augmented with uncertainty and trustworthiness models. (2) Continuously assess context relevant to the system requirements (in particular, determine the current risk) and correspondingly determine the requirements for the current situation. (3) Enable a continuous analysis of the compatibility of the current requirements and the current capabilities so that trustworthiness is guaranteed and performance can be optimized. (4) Collect in the field data during operation and thus enable continuous engineering and an optimally informed evolution of systems. To achieve these goals, CONTRAST brings together excellent partners from several countries which are amongst the leading institutions for relevant ingredients of the envisioned approach such as contract-based design, dynamic certification and analysis of AI-components. CONTRAST employs a use case driven approach with use cases from different industrial application domains (automotive, industrial automation, medical devices). Each of these use cases emphasizes specific aspects of the approach. This variety of use cases ensures that the whole approach is sufficiently addressed, is practicable in various domains and can be generalized. Moreover, based on the use cases we will illustrate the applicability of the concepts while at the same time increasing the competitiveness of the industrial partners in their respective markets. The consortium provides a good partner balance to cover the full innovation process, thus increasing European competitiveness in the area of trustworthy AI-based systems.

Project leader

Marc Zeller
Siemens AG, Germany
Project involvement CONTRAST
Alt Alt Alt Alt Alt Alt

Project publications