An Artificial Intelligence-Based Model-Driven Approach for Exposing Off-Nominal Behaviors

Abstract

With an increase in the automation of cyber-physical systems (e.g., automated vehicles and robots), quality problems such as off-nominal behaviors (ONBs) have also increased. While there are techniques that can find ONBs at the requirements engineering stage as it reduces the cost of addressing defects early in development, they do not meet the current industrial needs and often ignore functional safety. These techniques suffer from limitations such as scalability, need for significant human effort and inability to detect overlooked or unknown ONBs. To address these limitations we need a technique that analyzes requirements with respect to functional safety, but with less human effort. To achieve this, we propose our artificial intelligence-based model-driven methodology that provides a means to find ONBs during requirements engineering with minimal human effort. Our methodology utilizes existing approaches such as causal component model (CCM) and systems theoretic process analysis (STPA). We describe the details of each step of our approach and how our approach would support finding ONBs. Using our research and the results of our studies, we intend to provide empirical evidence that considering ONBs during requirements engineering stage and analyzing requirements with respect to functional safety can help create more robust designs and higher-quality products.